Exploring the Role of Generative AI for
Improving and Optimizing Sprint
Planning in Agile Development
University of Turku
Department of Computing
Master of Science (Tech) Thesis
Software Engineering
July 2025
Md Sharifur Rahman
Supervisors:
Antero Järvi (University of Turku)
Oshani Weerakoon (University of Turku)
The originality of this thesis has been checked in accordance with the University of Turku quality assurance
system using the Turnitin OriginalityCheck service.
UNIVERSITY OF TURKU
Department of Computing
Md Sharifur Rahman: Exploring the Role of Generative AI for Improving and
Optimizing Sprint Planning in Agile Development
Master of Science (Tech) Thesis, 76 p., 7 app. p.
Software Engineering
July 2025
Generative AI is transforming how software teams approach project management.
This thesis investigates the application of large language models, such as ChatGPT,
in Agile sprint planning. It begins with a review of current research and identi-
fies persistent challenges in traditional sprint planning, including subjective backlog
selection, inconsistent story point estimation, and uneven task assignment.
To address these issues, a web-based tool called GenSP was developed. GenSP
leverages GenAI APIs to generate sprint backlogs, estimate story points, break
down user stories, and assign tasks. The tool was evaluated using real project data,
and its effectiveness was further assessed through a survey of experienced Agile
practitioners.
The results indicate that GenAI can enhance sprint planning by automating backlog
refinement, story point estimation, and task breakdown. However, the study also
highlights concerns regarding the need for human oversight, the handling of complex
business logic, and the protection of sensitive data.
In summary, this research demonstrates that GenAI can make sprint planning more
effective and efficient. The findings provide practical guidance for teams considering
AI integration in Agile workflows and suggest directions for future research and
development.
Keywords: Artificial Intelligence, Agile Methodology, Scrum, Generative AI, Sprint
Planning, Project Management
Acknowledgement
I want to express my sincere gratitude to my supervisors, Antero Järvi and Oshani
Weerakoon, for their valuable guidance, continuous support, and encouragement
throughout this thesis journey. Their insightful feedback helped shape my research
and significantly improved the quality of my work.
I am especially grateful to the University of Turku (Software Engineering) for pro-
viding access to the ChatGPT API, which was instrumental in developing the tool
used in this research.
I am also thankful to all the participants who took the time to respond to my sur-
vey. Their experiences and thoughtful insights greatly contributed to this thesis’s
results.
Finally, I warmly thank my family and friends for their constant encouragement,
patience, and unwavering support, which made this accomplishment possible.
Contents
Acknowledgements 1
1 Introduction 1
1.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Goals and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Use of Generative AI . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Background 8
2.1 Fundamentals of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Overview Generative AI . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Generative AI in Software Engineering . . . . . . . . . . . . . . . . . 10
2.3.1 Code Development and Assistance . . . . . . . . . . . . . . . 11
2.3.2 Design and Creativity . . . . . . . . . . . . . . . . . . . . . . 11
2.3.3 Software Testing and Quality . . . . . . . . . . . . . . . . . . 12
2.3.4 Requirements Engineering . . . . . . . . . . . . . . . . . . . . 12
2.3.5 Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Agile Methodologies and Scrum . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Agile Methodologies Overview . . . . . . . . . . . . . . . . . . 14
CONTENTS 3
2.4.2 Scrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.3 The Significance of Scrum in Software Development . . . . . . 18
2.5 Limitations and Risks of Generative AI . . . . . . . . . . . . . . . . . 18
3 Sprint Planning and Generative AI 20
3.1 Sprint Planning in Scrum . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Challenges in Traditional Sprint Planning . . . . . . . . . . . . . . . 21
3.3 Current Research on Generative AI in Sprint Planning . . . . . . . . 22
3.4 Research Gap and Thesis Motivation . . . . . . . . . . . . . . . . . . 26
4 Methodology 27
4.1 Research Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1.1 Background Studies . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1.2 Prototype Design and Development . . . . . . . . . . . . . . . 28
4.1.3 Data Collection and Evaluation Strategy . . . . . . . . . . . . 29
4.1.4 Research Questions and Mapping to Methods . . . . . . . . . 30
5 System Design and Implementation 33
5.1 Core Functionalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.1.1 Projects Management . . . . . . . . . . . . . . . . . . . . . . . 34
5.1.2 Backlog Management . . . . . . . . . . . . . . . . . . . . . . 35
5.1.3 Story Points Estimation . . . . . . . . . . . . . . . . . . . . . 36
5.1.4 Sprint Generation . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1.5 Task Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1.6 Task Assignment . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1.7 User Authentication and Management . . . . . . . . . . . . . 38
5.2 GenSP System Architecture . . . . . . . . . . . . . . . . . . . . . . . 38
5.2.1 Backend Implementation . . . . . . . . . . . . . . . . . . . . . 39
5.2.2 Technologies and Tools . . . . . . . . . . . . . . . . . . . . . . 39
CONTENTS 4
5.2.3 ChatGPT’s API Integration . . . . . . . . . . . . . . . . . . . 41
5.2.4 Frontend Implementation . . . . . . . . . . . . . . . . . . . . 43
5.3 Prompt Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4 Challenges and Lessons Learned . . . . . . . . . . . . . . . . . . . . . 47
6 Results and Discussion 49
6.1 Evaluation using MyFlavoria Project . . . . . . . . . . . . . . . . . . 49
6.1.1 User Story Breakdown and Estimation . . . . . . . . . . . . . 50
6.1.2 Sprint Backlog Generation and Task Distribution . . . . . . . 54
6.1.3 Overall analysis . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.2 Survey Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.2.1 Survey Demographics . . . . . . . . . . . . . . . . . . . . . . . 59
6.2.2 Quantitative Results . . . . . . . . . . . . . . . . . . . . . . . 61
6.2.3 Qualitative Results . . . . . . . . . . . . . . . . . . . . . . . . 67
6.3 Mapping Research Questions with Results . . . . . . . . . . . . . . . 70
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.4.1 Evaluating GenSP Tool Data . . . . . . . . . . . . . . . . . . 72
6.4.2 Interpretation of Survey Results . . . . . . . . . . . . . . . . . 73
6.4.3 Key Contribution . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.4.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7 Conclusion 75
7.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
References 77
Appendices
A Survey Questions A-1
A.1 Generative AI in Sprint Planning: Survey Questions . . . . . . . . . . A-1
List of Figures
2.1 Relation among AI, ML, DL and GenAI [18]. . . . . . . . . . . . . . 10
5.1 GenSP Product Backlog. . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2 GenSP Story Point Generation. . . . . . . . . . . . . . . . . . . . . . 36
5.3 GenSP System Architecture. . . . . . . . . . . . . . . . . . . . . . . 39
6.1 GenAI Benefits in Sprint Planning (Survey Results) . . . . . . . . . 67
List of Tables
6.1 User Stories, Subtasks, and Estimated Points . . . . . . . . . . . . . . 50
6.2 GenSP Suggested Sprint and Task Distribution . . . . . . . . . . . . 54
6.3 Participant’s current company location . . . . . . . . . . . . . . . . . 59
6.4 Participant’s software development experience . . . . . . . . . . . . . 60
6.5 Familiarity with Scrum . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.6 Survey Results on GenAI in Sprint Backlog Generation . . . . . . . . 62
6.7 Survey Results on GenAI in Story Points Estimation . . . . . . . . . 63
6.8 Survey Results on GenAI in Story/Task Breakdown . . . . . . . . . . 65
6.9 Survey Results on GenAI in Task Distribution . . . . . . . . . . . . 66
1 Introduction
Artificial Intelligence (AI) is rapidly reshaping modern software engineering, driven
significantly by recent advancements in Generative AI (GenAI). Large language
models (LLMs) like ChatGPT, developed by OpenAI, have recently attracted sig-
nificant attention due to their impressive performance on various tasks, gaining
immense popularity among early adopters—some of whom describe them as disrup-
tive technologies across many domains [1]. Contemporary GenAI tools now support
software developers in code generation, suggesting optimal code refactoring [2], au-
tomatically generating and validating test cases [3], and assisting in documentation
[4]. Given these capabilities, new opportunities exist in exploring the integration of
GenAI within Agile methodologies, aiming to improve the efficiency, accuracy, and
overall effectiveness of Agile practices.
Agile software development has gained widespread industry adoption in response
to the demand for faster, cost-effective solutions in dynamic environments, with most
companies opting for agile due to its highly result-oriented approach [5]. So, differ-
ent Agile methodologies like Scrum and Kanban have become a standard approach
in software industries. Scrum is the most widely adopted Agile methodology, used
not only in software development but also in various other domains [6]. It struc-
tures development into short, focused cycles called sprints, which produce working
software features at regular intervals. This enables teams to incorporate feedback
and adjust to evolving requirements rapidly.
1.1 PROBLEM STATEMENT 2
Sprint planning, as described in the Scrum Guide, is a collaborative activity
that initiates each sprint by bringing the entire Scrum Team together to lay out
the work to be performed [7]. During this session, the team reviews and selects
the most relevant user stories from the product backlog based on the sprint goals.
The selected items are then estimated in terms of complexity and effort, often using
a point-based system. To ensure clarity and effective execution, these stories are
further broken down into smaller, actionable tasks. Finally, tasks are distributed
among team members according to their expertise, availability, and capacity. The
sprint planning ensures everyone is aligned on what can be delivered and how, with
input from the product owner, developers, and scrum master, and is timeboxed
based on sprint length to maintain focus and efficiency [8].
1.1 Problem Statement
In Scrum, Sprint Planning is a crucial team-oriented activity that defines the objec-
tives, scope, and deliverables for each development iteration [9]. Effective planning
is vital for the success of any software project. However, despite its significance,
traditional sprint planning methods face several inherent challenges [10] [11] [12]
[13].
One major issue is the heavy reliance on subjective and heuristic estimates from
team members, which frequently leads to inconsistencies and inefficiencies [10] [11].
Bias and subjectivity in these estimates can result in unrealistic or unclear sprint
goals and inaccurate effort estimation. Such problems often cause overcommitment
or underutilized team capacity, both of which are primary contributors to sprint
failures and diminished business value [12].
Moreover, traditional approaches often do not provide systematic methods for
identifying risks and dependencies early in the planning process [11]. When user
stories are manually broken down into tasks, important subtasks or dependencies
1.3 RESEARCH QUESTIONS 3
can be overlooked, leading to bottlenecks. Additionally, errors in task assignment
may produce uneven workloads, where some team members are overloaded while
others are underutilized. Sequential task dependencies further contribute to idle
time, which becomes especially inefficient in larger projects [13].
1.2 Goals and Objectives
This thesis investigates how GenAI can enhance sprint planning by addressing key
limitations found in traditional methods. The research specifically evaluates the im-
pact of GenAI across multiple, interconnected aspects of the sprint planning process.
The main objectives are as follows:
1. The effectiveness of GenAI in sprint backlog creation with user stories align
with sprint goals
2. The usefulness of GenAI in estimating user story points and supporting team
consensus during sprint planning
3. The effectiveness of GenAI in breaking down user stories into actionable tasks
4. The helpfulness of GenAI in suggesting skill-based task assignments and bal-
ancing team workloads during sprint planning
5. The overall GenAI’s impact on the effectiveness, efficiency, and quality of
sprint planning.
1.3 Research Questions
To address these objectives comprehensively, this thesis is guided by the following
research questions:
Research Question 1 (RQ1) : What is the value and effectiveness of GenAI in
1.4 METHODOLOGIES 4
sprint backlog creation, specifically in suggesting relevant user stories and identify-
ing missing details or dependencies?
Research Question 2 (RQ2): How effective is GenAI in estimating user story
points during sprint planning in terms of accuracy, facilitating team consensus, and
serving as a useful starting point for team discussions compared to traditional esti-
mation methods?
Research Question 3 (RQ3): How effective is GenAI in breaking down user
stories into actionable tasks, identifying task dependencies, and providing a useful
starting point for task refinement in sprint planning?
Research Question 4 (RQ4): How helpful is GenAI in suggesting task assign-
ments based on individual skills, experience, or development goals, and in balancing
workloads among team members during sprint planning?
Research Question 5 (RQ5): How does GenAI contribute to improving the over-
all effectiveness, efficiency, and quality of the sprint planning process?
1.4 Methodologies
This thesis adopts a mixed-methods approach, combining design science research
(DSR) [14] with empirical survey [15] analysis to investigate the potential of GenAI
in Agile sprint planning. The research is driven by two main focus areas:
1. Developing a web-based tool that leverages GenAI for key sprint planning
tasks
2. Evaluating its practical value based on feedback from Agile practitioners.
The research process began with a comprehensive literature review. This review
helped identify persistent challenges in traditional Agile sprint planning, such as
difficulties in aligning sprint backlog items with defined goals, subjective and in-
consistent effort estimation, and issues with workload distribution. Insights from
1.4 METHODOLOGIES 5
the literature also highlighted a gap in practical research on integrating GenAI into
sprint planning, motivating the development of a novel AI-supported solution.
To address these challenges, a functional prototype named GenSP is designed
and developed. This tool supports several core sprint planning activities: generat-
ing sprint backlogs that align with sprint goals and team capacity, estimating story
points from user story descriptions, breaking down user stories into actionable tasks,
and distributing tasks among team members. GenSP is built as a full-stack web ap-
plication, using React.js for the frontend, Express.js for the backend, and MongoDB
for data storage. The tool integrates OpenAI’s GPT-4 model as the core GenAI
engine, with prompt engineering techniques such as Chain-of-Thought (CoT) [16]
used to encourage logical, well-structured responses.
To evaluate the tool, its features were tested using the backlog of a real-world
software project. The focus was on assessing the completeness, relevance, and align-
ment of the AI-generated outputs with Agile practices. At this time it was not
feasible to involve a live Scrum team; this demonstration was complemented by a
structured survey distributed to experienced software developers working with Agile
methodologies. The survey included Likert-scale questions to assess perceived use-
fulness and accuracy, scenario-based items for direct feedback on AI outputs, and
open-ended questions for qualitative insights.
Before participating in the survey, respondents were introduced to the GenSP
tool through a video walkthrough or live demo. This ensured all participants had a
consistent understanding of the tool’s purpose and features, making their responses
more meaningful for analysis.
Overall, this methodology provides a rigorous foundation for evaluating how
GenAI can improve sprint planning, balancing theoretical research with practical
implementation and real-world feedback.
1.6 THESIS STRUCTURE 6
1.5 Use of Generative AI
In this thesis, GenAI was used for limited and specific editorial purposes. Ope-
nAI’s ChatGPT provided strictly grammatical and structural assistance, improving
sentence clarity, correcting language errors, enhancing document flow, and refining
academic tone. GenAI-processed content underwent careful verification to ensure
originality, coherence, and strict adherence to scholarly standards.
AI was not used in core intellectual contributions, including research design,
literature analysis, prototype development, evaluation methodology, and results in-
terpretation.
1.6 Thesis Structure
This thesis is organized into six chapters, each designed to address specific aspects
of the research objectives and provide a logical flow of discussion:
Chapter 2: Background - This chapter reviews the basics of AI and discusses
how GenAI has developed and where it is applied. The chapter also covers software
project management, with a focus on Agile methods like Scrum, and points out the
risks and limitations of using GenAI.
Chapter 3: Sprint Planning and Generative AI - This chapter explains
how sprint planning works in Scrum. It outlines problems in traditional planning and
examines how GenAI can help. The chapter also reviews recent research, highlights
the research gap, and explains the motivation for this thesis.
Chapter 4: Methodology - The methodology chapter describes how the re-
search was designed and carried out. It explains the use of design science research,
the development of the GenSP prototype, and how data was collected and evaluated.
The chapter also shows how each research question relates to the chosen methods.
Chapter 5: System Design and Implementation - This chapter details how
1.6 THESIS STRUCTURE 7
the GenSP tool was built. It covers its main features, including backlog manage-
ment, story point estimation, and task assignment. The technical stack, backend
and frontend design, and integration with GenAI are described. Challenges during
development and key lessons learned are also discussed.
Chapter 6: Results and Discussion - This chapter presents the results of
testing GenSP with real project data and survey feedback from Agile professionals.
It includes both quantitative and qualitative findings. The impact of GenAI on
sprint planning is analyzed, the research questions are answered, and study limita-
tions are discussed.
Chapter 7: Conclusion - The final chapter summarizes the main findings and
contributions of the thesis. It reflects on what the research means for practice and
suggests areas for future work.
This structure guides readers step by step, from the initial problem to the final
conclusions, and shows how GenAI can support better sprint planning in Agile
teams.
2 Background
In recent years, GenAI has become an important and influential force in software
engineering. AI-powered tools like GitHub Copilot are reshaping software devel-
opment by offering real-time code suggestions that boost productivity and reduce
cognitive load. [17]. These tools showcase how GenAI can be applied to real de-
velopment work and highlight its potential to support many stages of the software
engineering process, particularly in Agile methodology.
This chapter provides the foundation for understanding GenAI’s impact on soft-
ware engineering and Agile practices. It introduces core concepts in AI, Machine
Learning (ML), and Deep Learning (DL), then explores the growing role of GenAI
in software tasks such as coding, design, testing, and documentation. The chapter
also covers Agile methodologies, with a focus on Scrum and the challenges of sprint
planning.
2.1 Fundamentals of AI
AI is a branch of computer science that seeks to understand and replicate intelligence
by drawing on concepts from logic, mathematics, probability, learning, perception,
and action [18]. The field aims to develop autonomous systems capable of function-
ing in dynamic and uncertain environments [19].
The origins of AI trace back to the 1950s, when Alan Turing introduced the
idea of machine intelligence and proposed the "Turing Test" as a way to evaluate
2.2 OVERVIEW GENERATIVE AI 9
it [18]. Since then, advances in technology and data have contributed to the rapid
development of AI, which is now applied in areas such as language processing, image
recognition, autonomous vehicles, and healthcare.
ML, a key area within AI, enables systems to learn from data and adapt without
explicit programming [20]. Reviews such as that by Mian et al. [21] describe how
ML is applied in healthcare for disease diagnosis and personalized treatments, in
e-commerce for chatbots and recommendations, in finance for fraud detection, and
in sentiment analysis of social media and text data.
DL, a specialized subset of ML that uses Artificial Neural Networks (ANNs), is
fundamental to many modern AI systems, and it is commonly used in image and
speech recognition, Natural Language Processing (NLP), and autonomous driving
[20]. Surveys have documented DL’s diverse applications, including automating
medical diagnostics, improving energy-load forecasting, enhancing text classification
in NLP, signal de-noising, and computer vision tasks. Challenges related to limited
data and computational requirements are also noted in the literature [22].
2.2 Overview Generative AI
GenAI refers to systems that use generative models to create new text, images,
or other media by learning patterns from existing data [23]. In contrast to tradi-
tional AI, which primarily focuses on pattern recognition and predictive analysis,
GenAI models are capable of creating original outputs that often mirror work pro-
duced by humans. Major tech companies like Google, Meta, Microsoft, and OpenAI
have developed advanced generative AI technologies, with OpenAI notably releasing
DALL-E, a model capable of creating images from text descriptions using a variant
of the GPT-3 architecture [24]. GenAI systems are not limited to creative tasks;
they are increasingly being used as intelligent assistants that support knowledge
work and answer complex questions [25], thereby reshaping how both individuals
2.3 GENERATIVE AI IN SOFTWARE ENGINEERING 10
and organizations approach problem-solving and productivity.
GenAI relies on advanced DL architectures, primarily Deep Generative Models
(DGMs) such as Transformers, Latent Diffusion Models, VAEs, and GANs, each
offering distinct approaches to modeling and generating data [26]. Recent advances,
including multimodal integration and scalable Transformers, have enabled appli-
cations like text-to-image synthesis, though challenges like fuzzy VAEs and mode
collapse in GANs remain [27].
Figure 2.1: Relation among AI, ML, DL and GenAI [18].
2.3 Generative AI in Software Engineering
GenAI is rapidly gaining traction in software engineering. According to the Devel-
opers Survey of Stack Overflow 2024, 76% of respondents use or plan to employ AI
tools in their development processes [28]. This suggests that the use of GenAI tech-
nologies by software professionals is expanding. Ebert et al. [4] provides a practical,
practitioner-focused overview of how GenAI impacts software engineering, and it is
2.3 GENERATIVE AI IN SOFTWARE ENGINEERING 11
discussed here.
2.3.1 Code Development and Assistance
GenAI tools are increasingly vital in software engineering, automating code gener-
ation for a range of development tasks [4]. Popular AI-based code assistants such
as GitHub Copilot, Tabnine, ChatGPT, and Google Bard leverage ML and NLP to
provide real-time code suggestions and method implementations, offering valuable
support to developers, though they rarely produce fully correct, ready-to-use code
[2].
Evidence suggests that GenAI significantly enhances developer productivity. For
instance, developers using GitHub Copilot completed an HTTP server task 55.8%
faster than those without GenAI assistance [29]. Advanced models like GPT-4 can
generate entire functions or classes from natural language prompts and solve coding
problems from docstrings at or above human-level performance [4].
Modernizing legacy rule-checking systems is essential because they rely on man-
ual expertise, lack adaptability, cannot express complex logic, and are limited to
handling only simple, explicit requirements [30]. In addition, manual modernization
is slow, error-prone, and loses logic intent [31]. A study shows that GenAI mod-
ernizes legacy systems by automating manual tasks, handling ambiguous logic, and
enabling scalable, executable code generation [30].
2.3.2 Design and Creativity
GenAI is now making a noticeable impact in the creative and design areas of software
engineering. GenAI-powered design tools can automate tasks, boost creative idea
generation, and produce more unconventional outputs [32].
GenAI also supports more technical design tasks, such as chip layout and archi-
tectural planning. For example, in chip design, AI tools suggest circuit arrangements
2.3 GENERATIVE AI IN SOFTWARE ENGINEERING 12
that can boost performance [4]. In architecture, they help propose different build-
ing layouts for review. Overall, GenAI not only saves time but also encourages
innovation by offering fresh design options and efficient solutions [4].
2.3.3 Software Testing and Quality
Software testing and quality assurance are vital steps in software development. Re-
cent advances in GenAI, especially LLMs, enable dynamic test case generation,
vulnerability prediction, adaptation to software changes, and reduced reliance on
human testers [33]. This indicates that AI tools can read requirement documents or
user stories and quickly turn them into test cases. This approach helps cover more
scenarios and reduces the risk of missing important tests.
GenAI can automate the generation of unit and system test cases, craft test
assertions, visualize call graphs, execute and report tests, analyze bugs, support
debugging and automated repair, and document testing rationale [3]. By running
these tests automatically, AI tools help teams find problems faster and keep the
software reliable. GenAI can produce synthetic data for rare or edge cases. It
helps test and validate autonomous systems in fields like medicine and aerospace
[4]. Overall, GenAI makes testing more thorough and efficient, saving both time
and effort.
2.3.4 Requirements Engineering
GenAI has begun to play a valuable role in improving the requirements engineering
process, which is a critical early stage in software development. LLMs can auto-
mate and improve requirements engineering tasks by processing natural language,
generating documentation, and simulating stakeholder perspectives [34]. This also
indicates that GenAI-powered NLP tools can analyze requirements written in natu-
ral language to identify ambiguities, missing details, or even conflicting requirements
2.4 AGILE METHODOLOGIES AND SCRUM 13
within project documentation. By flagging vague or inconsistent statements, these
systems help teams clarify expectations early on, which minimizes misunderstand-
ings and reduces costly changes later in the project.
Notably, recent research [35] shows work on automating requirements engineering
in agile development using LLMs explores the use of GenAI on a larger scale. The
paper addresses the labor-intensive process of converting requirements documents
into user stories and test cases, proposing the tool GeneUS with the novel Refine
and Thought (RaT) prompting technique. The study positions itself as the first
implementation-level effort to automate user story generation with integrated test
specifications [35].
2.3.5 Documentation
A major use of GenAI is generating, modifying, and improving text. This is espe-
cially useful for managing project documentation. GenAI tools can quickly summa-
rize long documents and meeting notes, helping teams grasp important details and
focus on higher-priority tasks [4]. Many older codebases lack clear documentation,
but GenAI can help fill these gaps by creating documentation for legacy projects.
Automated documentation tools powered by AI also make it easier to keep project
records current. This not only improves code readability for the team but also helps
new developers understand the system more quickly. Overall, the integration of
GenAI in documentation ensures that knowledge is accessible, accurate, and easy
to maintain throughout the development lifecycle.
2.4 Agile Methodologies and Scrum
Software project management guides the entire development process, ensuring effi-
cient use of time, money, and resources. The following sections explore agile method-
2.4 AGILE METHODOLOGIES AND SCRUM 14
ologies in more detail, with a particular focus on Scrum.
2.4.1 Agile Methodologies Overview
Some traditional software development models rely on strict planning and lengthy
development cycles. These rigid approaches often struggle when customer require-
ments change frequently. In contrast, Agile methodologies offer flexibility to adapt
quickly to changing requirements. Agile methods effectively handle uncertainty by
emphasizing the skills of team members and the importance of their collaboration
in software development [36]
The foundation of Agile lies in the Agile Manifesto, and it was introduced in
2001 by a group of software professionals seeking to improve development processes
[37].
• Customer satisfaction through early, continuous delivery
• Embrace changing requirements
• Deliver working software frequently
• Business and developers collaborate daily
• Build projects around motivated people
• Communicate face-to-face
• Working software is the main progress measure
• Sustainable development pace
• Continuous attention to technical excellence
• Simplicity—maximize work not done
• Self-organizing teams produce best results
2.4 AGILE METHODOLOGIES AND SCRUM 15
• Regular reflection and adjustment
Agile methodology organizes work into short, repeatable cycles known as itera-
tions or sprints. Within each iteration, the team plans, develops, tests, and delivers
a working increment of the product. Customer and stakeholder feedback obtained
regularly is then used to refine goals and priorities for future iterations. Effective
communication is critical to success, and agile methods cannot thrive without a
foundation of strong teamwork and collaboration [38].
Agile methodology provides a flexible, people-focused approach to managing soft-
ware projects. It emphasizes close teamwork, continuous improvement, and regu-
lar stakeholder involvement. Several Agile frameworks, such as Scrum, Kanban,
and Scaled Agile Framework (SAFe), help teams apply these principles in prac-
tice. Scrum is the most widely used agile methodology, valued for its simplicity
and adaptability across diverse contexts [6]. It uses short, fixed-length cycles called
sprints and involves clearly defined roles and regular ceremonies like planning ses-
sions and retrospectives. Kanban is an agile method initially developed by Toyota
in the 1950s, emphasizing efficiency through visual workflow management, limiting
work-in-progress (WIP), and managing flow [39]. SAFe scales agile practices for
large enterprises and is structured around four levels—Team, Program, Portfolio,
and Value Stream—featuring Agile Release Trains, Program Increments, specialized
roles, and core values of quality, transparency, alignment, and program execution
[40].
Every Agile framework focuses on collaboration, adaptability to changing re-
quirements, and delivering incremental value. The choice among these frameworks
depends on organizational size, project characteristics, and team-specific needs.
2.4 AGILE METHODOLOGIES AND SCRUM 16
2.4.2 Scrum
Scrum is an agile framework that enables teams to work collaboratively, delivering
value incrementally through short cycles and continuous feedback [7]. It provides
a flexible structure that supports ongoing learning and improvement while allowing
teams to adapt practices to best fit their specific needs. [7].
Roles in Scrum are designed to ensure a balanced and self-organizing team struc-
ture. A Scrum Team is a small, cross-functional, and self-managing group consisting
of:
• Product Owner
• Scrum Master
• Developers
The Scrum Master, Product Owner, and Developers are focused on achieving
the product goal [9]. The product owner serves as the voice of the customer and
stakeholders, setting priorities in the product backlog so the team can focus on the
most valuable work. The Scrum Master guides the team, supports Scrum practices,
and helps solve problems that might slow down progress. The development team
includes professionals who plan, create, and test each part of the product during a
sprint. These teams bring together all the skills needed and are trusted to organize
and manage their own work.
Artifacts in Scrum provide transparency and help track progress. The artifacts
in Scrum are
• Product Backlog
• Sprint Backlog
• Increment
2.4 AGILE METHODOLOGIES AND SCRUM 17
The product backlog is a list of all desired features and changes for the product,
managed by the product owner. The Sprint Backlog contains the specific tasks and
user stories the team commits to complete in the current sprint. The increment
is the sum of all completed work at the end of a sprint, representing potentially
shippable functionality [9].
Scrum ceremonies are regular meetings that help organize the team’s work and
keep everyone on the same page. The Scrum ceremonies are [9]:
• The Sprint
• Sprint Planning
• Daily Scrum
• Sprint Review
• Sprint Retrospective
Sprint planning begins each sprint by letting the team choose what tasks to
tackle and how to approach them [7]. The daily scrum is a brief meeting where
team members update each other on progress, talk about any challenges, and plan
their next steps. At the end of the sprint, the Sprint Review allows the team to show
their finished work to stakeholders and gather feedback. The sprint retrospective
is a time for the team to look back on the sprint, discuss successes and areas for
improvement, and decide how to work better in the future [9].
Scrum’s roles, meetings, and tools help teams stay organized and flexible. This
structure makes it easier for teams to deliver value often, adjust to changes quickly,
and find ways to get better with each sprint.
2.5 LIMITATIONS AND RISKS OF GENERATIVE AI 18
2.4.3 The Significance of Scrum in Software Development
Scrum is now one of the most popular frameworks in the software industry, and it
is the most widely used agile framework, accounting for over half of all agile prac-
tices reported in use [41]. Its clear structure, adaptability, and focus on teamwork
make it a top choice for companies of all sizes, from startups to large corporations.
Surveys show that many software teams use Scrum and value its positive effects on
productivity, quality, and team morale.
A major reason for Scrum’s success is its ability to handle changing requirements
and customer feedback. By using short sprints and regular check-ins with stakehold-
ers, teams can deliver value quickly and adjust to new project needs. This approach
helps speed up delivery and lowers risks in complex projects. Scrum ceremonies
(e.g., sprint reviews) provide continuous stakeholder feedback, creating a powerful
way to develop products iteratively [6].
Scrum also encourages self-organizing teams and ongoing improvement. Develop-
ers begin each day with stand-up meetings to share updates and plan tasks. Rather
than having a project manager, Scrum teams are guided by a Scrum Master and
coordinate work through a Product Owner, who manages the backlog [6]. Many
organizations find that using Scrum improves communication, boosts employee en-
gagement, and supports better problem-solving. Its regular meetings and clear tools
make it easier for everyone to see progress and stay aligned [41].
2.5 Limitations and Risks of Generative AI
GenAI has rapidly advanced in recent years, enabling machines to create text, code,
images, music, and even videos with remarkable accuracy. However, despite their
potential benefits, GenAI systems come with significant risks and limitations.
Beltran et al. [42] explored that GenAI presents several risks for public sector
2.5 LIMITATIONS AND RISKS OF GENERATIVE AI 19
organizations. One of the main concerns is the potential leakage of confidential
information, as GenAI tools may unintentionally reveal sensitive government data.
Another significant risk is bias and discrimination, since these systems can carry over
unfair patterns from their training data, which may result in unjust decisions. GenAI
can also introduce cybersecurity weaknesses, making organizations more vulnerable
to data breaches [42].
Additionally, these tools are sometimes known to produce convincing but false
information, a problem called hallucination [43]. Other risks include high energy use,
which impacts the environment, and the possibility that heavy reliance on GenAI
could lead to reduced critical thinking and accountability within public sector teams
[42].
Over-dependency on GenAI can lead to diminishing creativity and ability to
generate original thoughts or designs [44]. The research of Abbas et al. (2024)
found that the use of GenAI tools like ChatGPT among university students brings
several risks to education. Frequent use is linked to increased procrastination, as
students may delay tasks when they rely on AI for support. Over time, depending
too much on these tools can weaken students’ memory and cognitive abilities [45].
3 Sprint Planning and Generative AI
This chapter highlights recent research on Agile sprint planning, identifies existing
gaps in the current literature, and explains the motivation for exploring the potential
of GenAI to address these challenges.
3.1 Sprint Planning in Scrum
Sprint planning is a key ceremony in Scrum that marks the start of each sprint.
During this meeting, the Scrum team collaboratively decides what tasks will be com-
pleted in the upcoming sprint and how to approach them [46]. The main purpose is
to create a clear, realistic plan and ensure everyone shares a common understanding
of the sprint’s goals.
All Scrum team members have distinct roles during sprint planning. The prod-
uct owner defines the sprint goal and relevant backlog items, while the Scrum team
selects tasks to complete and plans how to achieve them within the sprint [47]. The
product owner prioritizes and clearly explains the items from the product backlog,
and his/her objective is to add more value to the product in the current sprint.
The Scrum Master guides the meeting, helps resolve issues, and ensures the team
follows Scrum practices. The development team assesses tasks and breaks them into
manageable pieces. Agile teams use various estimation methods like Planning Poker
(Fibonacci-based cards), T-Shirt Sizing (abstract XS–XL sizes), Dot Voting (stake-
holder prioritization), Bucket System (grouped story sizes), Large/Uncertain/Small
3.2 CHALLENGES IN TRADITIONAL SPRINT PLANNING 21
(minimal categories), Ordering Method (collaborative ranking), and Divide Until
Max Size (splitting oversized stories) to assess effort and prioritize backlog items
efficiently [48]. the most widely used project management tools in Agile software
development include Jira, Trello, Asana, Microsoft VSTS, and Pivotal Tracker, due
to their strong support for Scrum and Kanban practices, customization, and team
collaboration features [49].
At the end of sprint planning, the result is a sprint backlog, which includes the
sprint goal, chosen items, and the plan to deliver them. Sprint planning is limited
in time, taking up to eight hours for a month-long sprint, and is shorter for shorter
sprints [46]. These outcomes provide a solid foundation, ensuring the team stays
organized, aligned, and focused throughout the sprint.
3.2 Challenges in Traditional Sprint Planning
Traditional sprint planning plays a central role in the Scrum framework, but it
frequently brings about a number of challenges for software development teams.
Biases and gaps in knowledge among team members can create various problems,
often making the sprint planning process less effective.
A typical product backlog is often large, complex, and constantly evolving. Tra-
ditional sprint planning often relies on subjective, heuristic estimates, causing in-
consistencies and inefficiencies [10] [11]. Setting a clear sprint goal is essential, as
studies show that teams without a well-defined objective struggle to stay focused
and achieve their targets [50]. Without careful planning, important backlog items
that are critical to the sprint goal may be overlooked, which can undermine the
success of the sprint
Estimation bias can lead to unclear goals, inaccurate effort predictions, and ei-
ther overcommitting or underutilizing team capacity [10] [11]. Common methods
like T-shirt sizing, Planning Poker, Dot-voting, and the Ordering Rule are popular
3.3 CURRENT RESEARCH ON GENERATIVE AI IN SPRINT PLANNING 22
among agile teams for estimating user story complexity [48]. While they are sim-
ple and foster collaboration, they remain subjective and prone to bias [10]. These
estimates are usually based on experience and gut feeling rather than data, which
can lead to either overcommitting or underutilizing team capacity. Teams often
struggle to accurately estimate the effort required for backlog items. Inexperience
with estimation techniques or having estimates imposed by management can under-
mine team self-organization. This often leads to unrealistic commitments, missed
deadlines, and increased technical debt [51].
Lack of systematic methods for identifying risks and dependencies, along with
manual task assignment, can lead to overlooked tasks, project bottlenecks, uneven
workloads, and increased idle time, especially in larger projects [11] [13].
Additionally, lack of shared understanding of project goals and priorities often
leads to misinterpretations, rework, and delays, especially in distributed teams where
ambiguous requirements cause teams to operate on differing assumptions [52].
Overall, these challenges make it difficult for teams to plan effectively and deliver
consistent value. Addressing these issues is essential for improving team performance
and achieving better results in Agile projects.
3.3 Current Research on Generative AI in Sprint
Planning
In recent years, research on GenAI has grown significantly within the fields of Agile
software development and Scrum. However, the specific application of GenAI to
sprint planning remains largely underexplored. There is considerable opportunity for
further investigation in this area, as only a limited number of studies have addressed
how GenAI can be used to improve and optimize the sprint planning process. This
gap highlights the need for more focused research to better understand and realize
3.3 CURRENT RESEARCH ON GENERATIVE AI IN SPRINT PLANNING 23
the potential benefits of GenAI in Agile sprint planning.
Prior research shows that it has been researched for some time to integrate AI
into the agile methodology. Dam et al. [53] propose that artificial intelligence can
fill this gap by automating tasks such as backlog refinement, sprint planning, effort
estimation, and risk management. Their framework outlines how AI engines can
analyze both structured and unstructured project data to support better decision-
making throughout Agile processes. However, they also emphasize that human
judgment and collaboration remain essential, and AI should be viewed as a tool to
support, not replace, Agile teams [53].
The research of Bahi et al. [54] highlights that GenAI technologies—such as
ChatGPT, GitHub Copilot, and Tabnine—are now helping Agile teams by au-
tomating tasks like code generation, testing, documentation, and backlog refine-
ment. GenAI also supports better decision-making, improves team collaboration,
and offers data-driven insights for sprint planning and progress tracking. While the
integration of GenAI can accelerate delivery, foster innovation, and improve software
quality, the paper also notes certain limitations. These include the need for human
oversight, ethical concerns, and the importance of context awareness. The authors
conclude that GenAI should be seen as a tool to support Agile teams rather than
replace them, and they encourage ongoing research and practical evaluation to make
the most of AI’s benefits while managing its risks [54].
Ostrowski et al. [55] highlighted that the specific integration of GenAI in IT
project management remains underexplored. GenAI offers significant potential
for automating routine tasks, optimizing resource allocation, and supporting data-
driven decision-making, but its adoption is challenged by issues such as data quality,
algorithmic bias, and organizational resistance. Despite proven benefits in other in-
dustries, there is a notable lack of empirical research and structured frameworks for
deploying GenAI within established project management methodologies like Agile
3.3 CURRENT RESEARCH ON GENERATIVE AI IN SPRINT PLANNING 24
and Scrum [55].
Several research studies have explored user story point estimation using GenAI.
Prior to the rise of GenAI, Choetkiertikul et al. [56] applied novel deep learn-
ing models to estimate user story points. The proposed models mark a significant
advance by automatically learning complex semantic patterns from raw user story
texts without the need for manual feature design. Their model, which combines Long
Short-Term Memory (LSTM) and Recurrent Highway Networks, outperforms estab-
lished baselines and alternative machine learning methods across a large and diverse
dataset, demonstrating the strong potential of deep neural networks to improve the
consistency and accuracy of story point estimation in Agile software development
[56].
Mallidi et al. [57] investigate how GenAI tools such as ChatGPT and GitHub
Copilot can improve Agile story point estimation and overall software development
productivity. Their findings show that GenAI tools can accelerate coding, enhance
code quality, and support multiple programming languages, resulting in 15–30%
time savings in various case studies. However, challenges remain around code con-
sistency, legal compliance, and potential over-dependence on AI. To address these
factors, the authors propose an adjustment model for story points that incorpo-
rates GenAI’s impact on efficiency and quality. They conclude that while GenAI
offers clear productivity benefits, teams should ensure proper training and careful
integration to mitigate associated risks [57].
Islam et al. [58] present a multimodal AI framework that integrates text, images,
and categorical data to enhance story point estimation in Agile software develop-
ment. Using models like BERT, CNN, and XGBoost, their approach predicts story
points with greater accuracy, especially for simpler tasks, achieving up to 77% ac-
curacy and a 0.84 F1-score without severity data. However, the study is limited
by a small, imbalanced dataset and noise in image inputs. The authors suggest
3.3 CURRENT RESEARCH ON GENERATIVE AI IN SPRINT PLANNING 25
that expanding the dataset, leveraging synthetic data, and fine-tuning models could
further improve performance and address current limitations [58].
There are several research studies on GenAI usage in Agile methodologies; no-
tably, Barcaui et al. [53] compared project plans made by a generative AI model and
by an experienced human project manager, using a real-world mobile app project
as a case study. Their results show that AI can quickly create clear and organized
project plans, especially for basic tasks like outlining scope, estimating costs, and
identifying risks. However, the plans created by the human project manager were
usually more detailed and realistic, especially when it came to understanding the
market, setting specific goals, planning resources, and dealing with complex issues.
The study found that AI and humans each have unique strengths: AI is fast and
consistent, but humans are better at creativity, contextual thinking, and handling
uncertainty. The authors conclude that the best results come when project man-
agers use AI as a tool, combining AI’s efficiency with human expertise to create
stronger, more practical plans [53].
Rahman et al. [35] work on automating requirements engineering, present a
tool named “GeneUS,” that leverages GPT-4 to automate the generation of user
stories and associated test case specifications from requirements documents. The
tool applies a novel prompt engineering technique called Refine and Thought (RaT),
which improves the quality and clarity of outputs by refining and structuring LLM
prompts. Through validation with software practitioners, the tool’s output was
found to be effective and highly rated in readability, understandability, and techni-
cal quality [35]. Adopting a similar approach, a tool to aid sprint planning could
use LLMs to automatically analyze backlog items and requirements, generate well-
formed user stories, estimate story points, and propose task breakdowns, thereby
reducing manual effort and enhancing the accuracy and efficiency of Agile planning
processes.
3.4 RESEARCH GAP AND THESIS MOTIVATION 26
The research described above highlights that only a limited number of studies
have directly addressed how GenAI can improve or optimize the sprint planning
process. Despite the growing interest in applying GenAI within Agile software de-
velopment, the use of GenAI in sprint planning remains largely underexplored.
3.4 Research Gap and Thesis Motivation
Recent research has looked at how GenAI can help in Agile software development
and Scrum. However, most of these studies focus on general tasks like backlog
refinement or story point estimation. Few have explored how GenAI can be used
for the full sprint planning process. There are still many open questions about how
GenAI can help teams choose backlog items, estimate effort, or spot risks during
sprint planning.
So far, most findings show that GenAI can speed up some Agile activities and
improve decision-making. But there is little evidence or clear frameworks for how to
use GenAI in real sprint planning meetings. This leaves a gap in our understanding
of how GenAI could make sprint planning more accurate and efficient.
This thesis is motivated by the need to fill that gap. It will investigate how
GenAI tools can support different steps of sprint planning and how teams can best
combine AI suggestions with their own experience. The aim is to provide practical
evidence and guidance for Agile teams who want to use GenAI to improve their
sprint planning outcomes.
4 Methodology
This chapter outlines the research methodology adopted in this thesis, which follows
a Design Science Research (DSR) approach [14] to develop and evaluate GenSP—a
GenAI-powered tool aimed at enhancing Agile sprint planning. GenSP supports
sprint backlog generation, story point estimation, task breakdown, and task distri-
bution, leveraging GPT-4 and prompt engineering. Its effectiveness was assessed
using historical sprint data and a structured survey of Agile practitioners.
4.1 Research Design
DSR is well-suited for research that involves creating and evaluating artifacts to
solve identified problems in practice. According to Hevner et al. [14], DSR aims
to enhance human and organizational capabilities by creating innovative artifacts
that provide solutions to real-world problems through iterative building and evalu-
ation. This approach aligns with the goals of this thesis. This research addresses
inefficiencies in traditional sprint planning by developing a novel GenAI-supported
tool.
4.1.1 Background Studies
The research began with an extensive literature review to identify persistent chal-
lenges in Agile sprint planning and the latest research on using GenAI in sprint
planning and relevant research gaps. As discussed in Chapter 3, common issues
4.1 RESEARCH DESIGN 28
include difficulties in aligning sprint backlog items with defined goals [50], as well as
challenges in accurately estimating effort due to subjectivity and bias [10]. These
problems have been shown to negatively affect team performance, reduce planning
accuracy, and undermine overall sprint effectiveness. The motivation for this study
arose from the growing potential of GenAI to address these inefficiencies by provid-
ing intelligent, data-driven support during the sprint planning process.
The primary objective of this thesis is to explore how GenAI can enhance the
sprint planning process by addressing key challenges faced in traditional Agile prac-
tices. To support this objective, a functional artifact was developed with core fea-
tures designed to assist critical sprint planning activities. The main functionalities
of the tool include,
• Generating sprint backlogs aligned with sprint goals and team capacity.
• Estimating story points from user story descriptions and acceptance criteria.
• Breaking down user stories into actionable tasks.
• Distributing tasks among team members based on their available capacity.
These objectives directly aligned with the research questions and aimed to ex-
plore how GenAI could assist in improving sprint planning accuracy and efficiency.
4.1.2 Prototype Design and Development
To meet the defined objectives, a functional prototype named GenSP is designed
and developed. It is a web application built using a modular, full-stack architecture
comprising React.js for the frontend, Express.js for the backend, and MongoDB
for data persistence. Each module in the system (e.g., projects, backlogs, sprints,
users) was designed with separation of concerns and scalability and integration with
existing project management tools in mind.
4.1 RESEARCH DESIGN 29
ChatGPT, developed by OpenAI, is a generative AI chatbot powered by LLMs,
capable of engaging in human-like conversations and performing a wide range of text-
based tasks [59]. In this thesis, ChatGPT (GPT-4) is used as the core GenAI engine
to support intelligent decision-making in sprint planning. To enhance the quality
and consistency of the AI’s responses, prompt engineering techniques—particularly
CoT prompting—were applied. CoT encourages the model to generate structured
and logical reasoning steps, improving the reliability of outputs in complex planning
scenarios [16].
The GenSP tool was developed to evaluate the feasibility of using GenAI to sup-
port specific sprint planning tasks, as described in Section 4.1.1. It serves as a proof
of concept to assess whether GenAI can provide meaningful assistance in generating
sprint backlogs, estimating effort, breaking down user stories, and distributing tasks
within Agile workflows.
4.1.3 Data Collection and Evaluation Strategy
In the current stage, it is not feasible to evaluate the GenSP tool with an actual
Scrum team during a live sprint cycle. Instead, the tool is tested using the backlog
of an existing software project—the MyFlavoria web application [60]. This applica-
tion’s product backlog and historical sprint data were utilized to examine whether
the tool’s core functionalities—namely sprint backlog generation, story point esti-
mation, task breakdown, and task distribution—could operate as intended. The
evaluation focused on the qualitative aspects of the AI-generated outputs, partic-
ularly their completeness, relevance to the sprint goal, and alignment with Agile
principles.
To supplement this demonstration, a structured survey was designed and dis-
tributed to software developers with practical experience in Agile methodologies.
According to Kasunic et al. [15], "A survey is a data-gathering and analysis ap-
4.1 RESEARCH DESIGN 30
proach in which respondents answer questions or respond to statements that were
developed in advance." This is the most valuable part of this thesis. The primary
aim was to gather their perceptions regarding the usefulness, accuracy, and reliabil-
ity of GenAI-powered sprint planning support, particularly across the four use cases
implemented in the GenSP tool.
The survey consisted of three main types of questions:
• Likert-scale questions to assess perceived usefulness, accuracy, and trust in
AI-generated outputs
• Scenario-based evaluation items, where participants reviewed and rated sample
outputs produced by the tool
• Open-ended questions, which invited respondents to share qualitative feed-
back, reflections, and concerns about GenAI in Agile sprint planning.
A Likert scale is made up of multiple Likert-type items, where respondents rate
their agreement on an ordered scale (e.g., Strongly Agree to Strongly Disagree);
scores are numerically coded, summed or averaged, and can be analyzed with both
parametric and non-parametric methods [61].
Before completing the survey, participants were introduced to the tool and its
capabilities through either a brief video walkthrough or a live demonstration session.
This ensured that all participants had a clear understanding of the tool’s purpose,
features, and user flow, allowing for more informed and meaningful responses.
4.1.4 Research Questions and Mapping to Methods
Five main research questions mentioned in section 1.3 guide this thesis, each focusing
on a key area where GenAI could support and enhance Agile sprint planning. The
approach to answering each question combines artifact demonstration, empirical
evaluation, and practitioner feedback to ensure technical and practical relevance.
4.1 RESEARCH DESIGN 31
To address RQ1, the GenSP tool was designed to automate sprint backlog cre-
ation by analyzing sprint goals, team capacity, and available backlog items. The ef-
fectiveness of these AI-generated suggestions was evaluated through both hands-on
demonstration using the Flavoria project’s sprint data and survey questions asking
practitioners to assess the value and relevance of the AI’s outputs.
For RQ2, GenSP generates story point estimates for user stories based on their
descriptions and acceptance criteria. These estimates were reviewed in the demon-
stration phase and rated by survey participants for perceived accuracy and useful-
ness, with additional feedback gathered on how GenAI may influence estimation
discussions within teams.
To investigate RQ3, the tool provides automatic task breakdowns for user sto-
ries and highlights potential dependencies. Both the demonstration and the survey
explored the clarity, completeness, and practicality of these task breakdowns, as
well as participant comfort in using GenAI outputs as a starting point for further
refinement.
RQ4 is examined by analyzing the tool’s capability to recommend task assign-
ments that account for team capacity and expertise. The demonstration compared
GenSP’s automated task assignments to historical team practices, while the sur-
vey collected practitioner opinions on the usefulness and fairness of AI-driven task
distribution.
This overarching question, RQ5 is answered through a combination of artifact
evaluation and practitioner feedback. The demonstration phase assessed GenSP’s
practical outcomes, while the survey gathered broader insights on how GenAI affects
planning quality, team satisfaction, and sprint outcomes.
In summary, each research question is addressed through a combination of tool-
based demonstration (aligned with the DSR methodology) and practitioner feedback
(via surveys). This dual approach ensures both practical validation and empirical
4.1 RESEARCH DESIGN 32
evaluation of GenAI’s role in Agile sprint planning.
5 System Design and
Implementation
This chapter presents the development process and architectural details of GenSP
which is an AI-powered application.
To build this application, several modern technologies were combined. The back-
end is built with Express.js, which handles all the server-side logic and data man-
agement. The frontend uses React.js to create a user-friendly interface where team
members can interact with sprint planning tools and view AI-generated suggestions.
MongoDB is used as the database to store all relevant project data, including prod-
uct backlogs, sprints, tasks, and user profiles. Most importantly, the application
is connected to the ChatGPT API, which allows the application to generate user
stories, estimate story points, and break down tasks.
This chapter details the architecture, design choices, and technical implementa-
tion of the system. It explains how each component was developed and integrated
to provide a seamless user experience and ensure the reliability and security of the
application. By documenting the development process, this chapter lays the foun-
dation for evaluating the practical benefits and limitations of GenAI in real-world
sprint planning scenarios.
5.1 CORE FUNCTIONALITIES 34
5.1 Core Functionalities
At the start of developing GenSP, the initial scope and requirements were fairly
modest. The main objective was to create basic REST (Representational State
Transfer) APIs to facilitate core sprint planning tasks like sprint backlog manage-
ment and story point estimation. A RESTful API is a standardized interface that
enables secure and efficient communication between computer systems over the in-
ternet [62]. However, as the project progressed, it became evident that these basic
features were not sufficient. The application needed additional capabilities to fully
support the complexities of GenAI-powered Agile project management. This real-
ization led to the inclusion of more features, expanding the system to better address
the needs of users and deliver a more comprehensive solution.
Ultimately, GenSP evolved into a comprehensive platform featuring several in-
terconnected backend modules, moving well beyond the initial vision of a simple
API. Each module was carefully designed to integrate with others, resulting in a
seamless user experience that supports a wide range of Agile project management
tasks. While this broader scope introduced additional complexity to both the back-
end and frontend, it was essential to deliver a tool that truly meets the needs of
Agile teams. As a result, GenSP now offers not only advanced GenAI-driven ca-
pabilities for sprint planning and suggestions but also a robust set of core project
management features. The following sections provide a detailed overview of these
key integrated functionalities.
5.1.1 Projects Management
In order to keep the product backlog separate, GenSP allows users. Each project
operates as an independent workspace and contains its own set of modules. For
example, every project maintains a unique product backlog consisting of multiple
backlog items, ensuring that work items from different projects remain organized and
5.1 CORE FUNCTIONALITIES 35
do not overlap. The project management module supports all basic CRUD (Create,
Read, Update, and Delete) operations, enabling users to easily add new projects,
view project details, modify information, and remove projects as needed. This
structure provides a clear and efficient starting point for managing Agile workflows
and keeps project data organized throughout the development process.
Figure 5.1: GenSP Product Backlog.
5.1.2 Backlog Management
The product backlog is a fundamental component of Scrum, and GenSP provides
robust features for managing backlog items through comprehensive CRUD opera-
tions. In the current version, users can create and manage different types of backlog
items, including epics, stories, bugs, and tasks. Each item can be assigned a priority
level—High, Medium, or Low—and its status can be updated to reflect its progress,
such as To Do, In Progress, or Done. This flexibility allows teams to adjust item
statuses as work advances during a sprint. Additionally, user story points can be en-
tered manually or automatically generated using GenAI, supporting more consistent
5.1 CORE FUNCTIONALITIES 36
and accurate effort estimation throughout the development process.
5.1.3 Story Points Estimation
When a backlog item is created in GenSP, there is an option to generate user story
points automatically. GenAI estimates the story points based on the item’s title
and description, including the acceptance criteria. Once generated, the estimated
story points are entered directly into the appropriate field. Users can also click
the “Details” button to view an explanation from GenAI, describing the reasoning
behind the assigned story points. This feature provides a valuable starting point for
team discussions, helping to save time and streamline the process of estimating user
story points.
Figure 5.2: GenSP Story Point Generation.
5.1 CORE FUNCTIONALITIES 37
5.1.4 Sprint Generation
GenSP enables GenAI-assisted sprint backlog generation to streamline the planning
process. Users begin by entering the sprint goal and specifying the sprint duration in
days (for example, entering “10” for a 10-day sprint). Next, the development team is
selected, and each member’s total available working hours for the sprint are provided
(e.g., 80 hours for a developer available all 10 days, assuming an 8-hour workday).
After entering these details, the user can click the “Generate Sprint” button. GenSP
will then utilize ChatGPT’s API to intelligently suggest sprint backlog items, taking
into account unfinished backlog items, the defined sprint goal, and the team’s total
capacity. This feature helps ensure that the generated sprint plan aligns closely with
project objectives and team availability.
5.1.5 Task Breakdown
GenSP also provides automatic user story breakdown, helping users divide a user
story into smaller, manageable tasks. When a user story point is generated, GenAI
simultaneously analyzes the story’s title and description to suggest a set of subtasks.
This breakdown not only clarifies the complexity and scope of work required to
complete the user story but also helps identify potential dependencies that might
otherwise be overlooked.
5.1.6 Task Assignment
GenSP also features intelligent task assignment during sprint backlog generation.
After receiving the sprint goal and team capacity, GenSP uses ChatGPT’s API to
recommend which backlog items to include and assigns them to the most suitable
team members based on their roles (e.g., frontend developer, backend developer,
UI designer) and their past tasks. While team preferences may still influence final
assignments, GenSP provides a draft plan that streamlines planning and supports
5.2 GENSP SYSTEM ARCHITECTURE 38
balanced workload distribution, making it easier for teams to review and adjust
during sprint planning.
5.1.7 User Authentication and Management
The system includes essential user management features, such as user authentica-
tion, to ensure security. For simplicity, only users with the admin role are permitted
to manage user accounts through CRUD operations. All other users can log in
using their email and designated password and may only modify their own infor-
mation. The role-based access controls are kept straightforward, as advanced user
permissions are not directly relevant to improving sprint planning. However, im-
plementing basic authentication helps protect the system from unauthorized access
while allowing legitimate users to securely use the application.
5.2 GenSP System Architecture
The GenSP application is designed with a modular architecture that integrates
four key components: the frontend, backend, database, and GPT API integration.
Figure 2.1 presents the high-level architecture of the GenSP application. This ar-
chitecture ensures a seamless flow of information between users and the system,
enabling efficient and AI-enhanced sprint planning. The frontend offers an intu-
itive user interface for managing all sprint planning activities, while the backend
handles business logic, data processing, and communication with both the database
and GPT API. The database maintains persistent and secure storage of all project
data. Integration with the GPT API empowers the system to provide intelligent
suggestions for user stories, estimations, and task breakdowns.
5.2 GENSP SYSTEM ARCHITECTURE 39
Figure 5.3: GenSP System Architecture.
5.2.1 Backend Implementation
The backend serves as the central logic and coordination layer of the GenSP appli-
cation. Its primary responsibility is to process incoming requests from the frontend,
execute core business logic, and manage seamless data exchange between the user in-
terface, the database, and external services such as the ChatGPT API. The backend
is structured around a set of RESTful endpoints, each designed to support specific
operations required for effective sprint planning and project management.
5.2.2 Technologies and Tools
The backend of GenSP acts as the central logic layer, processing frontend requests,
managing business logic, and coordinating data flow between the UI, database, and
external services like the ChatGPT API. It is organized around RESTful endpoints
tailored for sprint planning and project management.
Technologies and Tools
GenSP uses a modern, scalable stack:
• Express.js: This Node.js framework powers the RESTful API, supporting
5.2 GENSP SYSTEM ARCHITECTURE 40
scalable and robust backend logic [63].
• MongoDB:A flexible NoSQL database, well-suited for managing Agile project
data [64].
• Mongoose: ODM library providing schema-based data modeling and valida-
tion for MongoDB [65].
• OpenAI GPT API: Integrated via OpenAI’s Node.js SDK to enable GenAI-
powered sprint planning features [66].
Project Structure
GenSP adopts a modular, MVC-inspired architecture (excluding the traditional view
layer since the frontend consumes API endpoints). The service-repository pattern
enhances code maintainability and scalability [67]. Models define schemas, con-
trollers handle requests, services implement business logic, and repositories manage
database interactions.
• Route: Exposes RESTful endpoints for project operations.
• Middleware: Validates requests and manages permissions.
• Model: Defines project schemas in MongoDB via Mongoose.
• Controller: Processes project-related HTTP requests.
• Service: Handles business logic for project tasks.
• Repository: Abstracts MongoDB interactions.
Database Design
MongoDB’s document-based structure and Mongoose’s schemas provide flexible
and scalable data management, supporting clear relationships between entities like
5.2 GENSP SYSTEM ARCHITECTURE 41
projects, users, backlog items, sprints, and tasks. This approach ensures data in-
tegrity and efficient CRUD operations throughout the application.
Security and Best Practices
Security best practices include encrypted password storage, JWT-based authentica-
tion, and rigorous input validation. These measures protect user data and maintain
system integrity.
5.2.3 ChatGPT’s API Integration
A central feature of the GenSP application is its integration with OpenAI’s Chat-
GPT API to enable intelligent automation for sprint planning tasks. The backend
leverages LLM to generate user stories, estimate story points, break down tasks, and
assign responsibilities. This chapter details how the ChatGPT API is integrated, the
specific techniques used for prompt engineering (such as chain-of-thought prompt-
ing), how context is maintained across API calls, and the role of the Node.js OpenAI
SDK.
The GenSP system’s backend communicates with ChatGPT using the official
OpenAI Node.js SDK. This SDK makes it easier to send requests to OpenAI’s API
and get responses back. This way, GenAI features can be added smoothly to REST
endpoints. Important steps for this integration include setting up the system by
securely storing API keys. All interactions with the OpenAI API are handled in a
specific service module to keep things organized. The system also includes strong
error handling to catch problems with the API and provide helpful information to
the user if something goes wrong.
To begin, an OpenAI client must be created. This client acts as the connection
point to the OpenAI API. Once the client is set up, it can be used to send requests to
ChatGPT. This process typically involves specifying the model to use and providing
5.2 GENSP SYSTEM ARCHITECTURE 42
the user’s message. ChatGPT then processes this request and sends back a response.
The system receives this response from the API via the client.
// Load environment variables from .env file
dotenv.config();
const configuration = new Configuration({
apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
export default openai;
Before sending a request to the ChatGPT API, the system first selects the ap-
propriate model—“gpt-4” in this case—due to its advanced intelligence. The request
is then prepared by constructing a sequence of messages. This includes a system
message that sets the AI’s role, such as “You are an Agile Sprint Planner.” Finally,
the main prompt is tailored to the specific sprint planning task at hand, whether it
involves backlog generation, story point estimation, or task breakdown. A tempera-
ture parameter, often set to 0.5, is specified to balance creativity and predictability
in the AI’s responses. Once the messages and settings are assembled, the backend
sends the request to the OpenAI chat completions API [68] and waits for the AI-
generated response, which is then processed and integrated back into the application
workflow.
const response = await openai.chat.completions.create({
model: "gpt-4",
messages: [{
role: "assistant",
5.2 GENSP SYSTEM ARCHITECTURE 43
content: "You are an Agile Sprint Planner."
},
{role: "user", content: prompt}
],
temperature: 0.5,
});
5.2.4 Frontend Implementation
Initially, the scope of this project did not include a frontend implementation. How-
ever, it soon became clear that developing a frontend interface was crucial. The
frontend helps users clearly visualize how GenAI can enhance Agile sprint planning
by making it interactive and user-friendly. For this purpose, React.js, one of the
most popular and robust frontend frameworks, was chosen for development.
The frontend architecture of GenSP follows a modular structure, mirroring the
organization of backend modules such as Projects, Backlog Items, Sprint Genera-
tion, and User Management. Each frontend component corresponds to its respective
backend module, which ensures a consistent, clear, and maintainable codebase. This
modular design allows components to be developed, tested, and maintained inde-
pendently, facilitating efficient collaboration and scalability. Furthermore, aligning
frontend modules with backend logic simplifies data handling and enhances the over-
all coherence of the application, providing a seamless user experience.
For the GenSP frontend implementation, two key libraries were used: Axios
and Bootstrap. Axios was chosen as the primary HTTP client due to its simplicity,
efficient handling of API requests, built-in promise support, and clear error manage-
ment. It streamlined communication between the React frontend and the Express.js
backend by making asynchronous requests easy and maintainable. Bootstrap was
integrated to provide rapid development of the application’s styling and responsive
5.3 PROMPT ENGINEERING 44
layout. Its robust, grid-based system and pre-built components enabled quick cre-
ation of an intuitive and visually appealing user interface, ensuring compatibility
across various devices and screen sizes.
5.3 Prompt Engineering
A critical component of integrating GenAI into the GenSP application is design-
ing effective prompts for the ChatGPT API. Prompt engineering is the practice of
designing and structuring prompts to effectively communicate tasks to AI models
[69]. The quality and structure of these prompts directly influence the relevance
and usefulness of the AI-generated outputs. Its benefit is that well-crafted prompts
improve the relevance, accuracy, and fairness of AI outputs while enabling the model
to handle complex problems and reducing errors [69].
For each sprint planning task, prompts are carefully crafted to provide clear
instructions and guide the LLM through a logical reasoning process. CoT technique
is used where prompts include not only the question but also explicit, step-by-step
reasoning (rationales) leading to the answer, encouraging the model to show its
intermediate thought process before giving the final response [16].
For example, when estimating user story points, the prompt is constructed to
simulate the thought process of an experienced Agile team member. The prompt in-
structs the AI to analyze the user story by first identifying its key components—such
as user interface, backend, database, or API requirements. The AI is then guided to
break down the story into smaller subtasks, assess the complexity of each subtask,
and consider factors such as technical difficulty, dependencies, and potential risks.
To encourage thorough and systematic reasoning, the prompt uses a step-by-step
chain-of-thought approach. This includes explicit steps for understanding the story’s
scope, evaluating unknowns, and estimating the required effort using the Fibonacci
sequence—a standard practice in Agile estimation. The prompt also directs the AI
5.3 PROMPT ENGINEERING 45
to provide both a numeric estimate and a summary of the reasoning behind the
suggested story points, ensuring that the output is transparent and actionable for
the development team.
By employing detailed, task-specific prompts like the example below, GenSP
maximizes the value of GenAI in sprint planning. This structured prompt-building
process helps produce consistent, logical, and context-aware recommendations that
align with real-world Agile practices.
const prompt = `
Act as an Agile team member responsible for estimating story points
for a user story. Follow this step-by-step chain of thought to
estimate the story points based on the provided story title,
description, and acceptance criteria. Be detailed and logical
in your reasoning.
Understand the Scope:
Identify the key components of the story, such as UI, Backend,
Database, API, etc.
Break Down the Story into Tasks:
List the specific subtasks required to complete the story.
Assess the Complexity:
Evaluate the complexity of each subtask. Consider factors like
technical difficulty, dependencies, and effort required.
Consider Risks, Unknowns, and Dependencies:
Identify any potential risks, unknowns, or dependencies that
5.3 PROMPT ENGINEERING 46
could impact the effort required.
Estimate Story Points:
Based on the above analysis, suggest a story point estimate
using the Fibonacci sequence (1, 2, 3, 5, 8, 13, etc.).
Assume that 1 story point represents a day of work for an
experienced team member and all the members of the development
team is highly experienced and familiar with the project.
Final Estimate:
Provide the final story point estimate and a brief summary of the
reasoning. Use floor or ceiling numbers for the total story
points estimate so that it follows Fibonacci
sequence (1, 2, 3, 5, 8, 13, etc.).
Here is the user story title, description, and acceptance
criteria:
Title: ${storyTitle}
Description: ${storyDescription}
Now, follow the chain of thought and provide a detailed story
point estimate with proper reasoning.`;
To ensure ChatGPT provides relevant and connected responses, especially during
ongoing planning, the GenSP backend carefully builds a messages array for each API
5.4 CHALLENGES AND LESSONS LEARNED 47
request and stores it as chat history. For tasks that involve multiple exchanges, like
generating a sprint backlog, the conversation history is also included. This helps
the AI remember previous responses and user follow-ups, allowing it to learn from
how past sprints were managed, including story selection, effort estimations, task
breakdowns, and team assignments.
5.4 Challenges and Lessons Learned
The development of the GenSP application presented various technical and design
challenges that shaped the final outcome of the project. One of the most significant
challenges was the evolving scope of the system. Initially, the goal was to develop a
lightweight API focused solely on sprint planning and backlog management. How-
ever, as development progressed and real-world use cases were considered, it became
evident that broader project management capabilities were essential. This shift re-
quired the introduction of new features, such as multi-project support, team member
assignments, and user role management. As a result, the backend needed substan-
tial restructuring to accommodate these changes, including updates to data models,
API endpoints, and access control logic.
Integrating the GPT API posed another key challenge. Crafting prompts that
produced accurate and consistent results required multiple iterations. For tasks
like sprint backlog generation or story point estimation, prompts had to be specific
enough to guide the AI’s reasoning but concise enough to remain within token limits.
To maintain output reliability, a two-step prompt structure was adopted—first to
generate an explanation, then to request structured output. However, tuning this
interaction was not straightforward. A common difficulty involved balancing the
creativity and consistency of the AI’s responses, as large language models like GPT
are prone to variability.
Another observed issue was hallucination, where the AI produced plausible but
5.4 CHALLENGES AND LESSONS LEARNED 48
incorrect or exaggerated outputs. For example, when estimating story points, the
AI occasionally assigned high estimates if the story lacked a clear description or
defined acceptance criteria. This highlighted the importance of prompt clarity and
well-prepared input data.
From a technical standpoint, ensuring smooth communication between the back-
end and frontend was also a challenge. The dynamic nature of AI-generated sugges-
tions meant that API responses needed to be handled carefully, with proper state
management and real-time UI updates. Security was another important concern.
Protecting sensitive data, managing user access, and securing API keys required
dedicated attention throughout the implementation.
Several important lessons were learned during this process. First, project re-
quirements can change quickly, so it is important to design flexible systems that can
adapt over time. Successful integration of GenAI depends heavily on precise prompt
engineering and strong error handling. Incorporating AI into a traditional project
management system is not trivial—it involves managing dependencies between var-
ious modules while maintaining the logic and usability of the system. Involving
users early in testing can reveal valuable insights and usability issues that guide
further development. Finally, maintaining a clear separation of concerns across the
backend, AI logic, and frontend layers proved vital for scalability and long-term
maintainability.
Overall, these experiences underscored the importance of iterative development,
adaptability, and collaboration. Building an AI-assisted tool like GenSP required not
only technical skill but also a mindset open to learning and continuous refinement.
6 Results and Discussion
The evaluation in this thesis was conducted using two methods. First, the GenSP
tool was tested against completed sprint data from the MyFlavoria project backlog
to ensure its key features were assessed with realistic project information. Second,
a structured survey was carried out among experienced Agile practitioners and soft-
ware developers.
6.1 Evaluation using MyFlavoria Project
MyFlavoria [60] is an app that lets customers track their meals, view nutritional in-
take, and monitor biowaste at the Flavoria research restaurant, contributing to food
innovation and waste reduction research. To assess the effectiveness of GenSP’s
AI-assisted sprint planning, the tool was applied to a real-world scenario in the
MyFlavoria software project. Its product backlog, historical sprint data, and previ-
ous sprint goals provided a solid benchmark for comparison. The evaluation focuses
on how GenSP performed in user story breakdown, story point estimation, task
assignment, and overall sprint planning, relative to the project’s traditional sprint
history.
For this project, the team consisted of seven members: four software engineers
and three UI/UX designers. The software engineers formed the development team,
while the UI/UX designers made up the design team. The members are referred to
as follows:
6.1 EVALUATION USING MYFLAVORIA PROJECT 50
• Software Engineer 1 (SE1)
• Software Engineer 2 (SE2)
• Software Engineer 3 (SE3)
• Software Engineer 4 (SE4)
• UI/UX Designers 1 (UX1)
• UI/UX Designers 2 (UX2)
• UI/UX Designers 3 (UX3)
6.1.1 User Story Breakdown and Estimation
GenSP was used to analyze user stories from the MyFlavoria backlog. For each user
story, the tool generated a breakdown of actionable tasks and provided story point
estimates. Table 6.1 illustrates this process by mapping representative user stories
to their respective GenSP-generated tasks and point estimates:
Table 6.1: User Stories, Subtasks, and Estimated Points
User Story Broken Down Tasks Story Points
As a student, I want to see
how my lunch compares to
national dietary recommen-
dations, so that I can adjust
my diet to make it healthier.
- Design comparison UI
- fetch dietary API
- Integrate with user data
- Testing
8
6.1 EVALUATION USING MYFLAVORIA PROJECT 51
As a diner, I want to see to-
day’s menu so I can decide
what I want to eat
- Design today’s menu UI
- Fetch menu API
- Integrate frontend
- Test menu display
3
As a diner/allergic I want
to be able to inspect the in-
gredients of each food op-
tion (in the restaurant and
elsewhere) so I know what I
want to eat
- Add ingredient info to DB
- Ingredient info UI
- Modal/pop-up
- Test feature
5
As a diner I want to see the
week’s menu in the restau-
rant so I know what they’re
serving on other days
- Fetch weekly menu
- Week-view menu UI
- Integrate with frontend
- Testing
3
As a vegetarian/vegan, I
want to see the vegan food
option first in the menu so
I don’t have to go through
options not meant for me,
and as a meat eater, I want
to see the meat food options
first in the menu so I don’t
have to go through options
not meant for me
- Dietary preference settings
- Sorting/filter logic
- UI for ordering
- Test personalization
5
6.1 EVALUATION USING MYFLAVORIA PROJECT 52
As a diner I want to very
easily be able to view the
QR code needed at check-
out so that it doesn’t slow
my checkout
- QR access button UI
- Integrate QR generator
- Test QR display
3
As a diner, I want the
garbage point to be less
crowded so I can exit the
restaurant fast
- Analyze current process
- Design solution
- Implement notifications/alerts
- Test with users
8
As a diner, I want to know
when the QR code has
been successfully read at the
checkout so that I do not
have to guess if it has been
properly read or not.
- Feedback UI for scan
- Integrate with checkout
- Test response
2
As a restaurant owner I
want to inform possible cus-
tomers about the opening
hours and location of the
restaurant, so that cus-
tomers will find their way
to the restaurant and at the
correct time.
- Manage info API fetch
- Design info UI
- Test visibility
2
6.1 EVALUATION USING MYFLAVORIA PROJECT 53
As a diner I want to be able
to find the restaurant open
times easily so I know when
I can eat there
- Quick access UI
- Backend integration
- Test feature
2
As a restaurant employee
responsible for the food I
want to get as much cus-
tomer feedback as possible
so I can make changes to the
menu
- Feedback UI
- Feedback storage throuhg API
- Admin analytics view
- Test feedback process
5
As a new user I want a video
version of instructions so I
can easily understand how
the app and the restaurant
works
- Script and record video
- Integrate in onboarding
- Test accessibility
3
As a non-finnish-speaking
customer, I want to be able
to change the language of
the application so I can un-
derstand it
- Language selection UI
- Translate content
- Dynamic switching
- Test functionality
8
This approach demonstrated that GenSP could produce detailed, actionable task
breakdowns and realistic effort estimations, comparable to what experienced Scrum
teams deliver manually.
6.1 EVALUATION USING MYFLAVORIA PROJECT 54
6.1.2 Sprint Backlog Generation and Task Distribution
Beyond analyzing individual user stories, GenSP generated complete sprint plans
based on sprint goals, team capacity, and backlog status. It selected user stories,
defined sprint goals, and recommended balanced task assignments. These plans were
then compared to the MyFlavoria project’s historical sprints.
Table 6.2: GenSP Suggested Sprint and Task Distribu-
tion
Sprint Sprint Goal Selected User
Stories
Assigned Workers
1 Launch core
menu and essen-
tial info
2 (Today’s
menu), 4
(Week’s menu),
9 (Owner info),
10 (Open times)
SE1: Today’s menu API, SE2:
Week’s menu API, SE3: Inte-
gration, SE4: Owner/Open times
backend; UX1: Today’s menu UI,
UX2: Week’s menu UI, UX3:
Owner/Open times UI
2 Personalize
menu and en-
able QR basics
5 (Menu sort-
ing), 6 (QR
access), 8 (QR
scan feedback)
SE1: Sorting logic, SE2: QR in-
tegration, SE3: Profile settings,
SE4: Scan feedback logic; UX1:
Sorting UI, UX2: QR UI, UX3:
Scan feedback UI
Continued on next page
6.1 EVALUATION USING MYFLAVORIA PROJECT 55
Table 6.2 – continued from previous page
Sprint Sprint Goal Selected User
Stories
Assigned Workers
3 Dietary safety
and language
preparation
3 (Ingredients
inspection),
13 (Language
change - back-
end & UI setup)
SE1: Ingredients backend, SE2:
Language selection backend, SE3:
Ingredient modal, SE4: Content
translation; UX1: Ingredients UI,
UX2: Language UI, UX3: Test-
ing
4 Dietary compar-
ison & feedback
collection
1 (Dietary
comparison),
11 (Feedback
collection)
SE1: Dietary backend, SE2:
Feedback backend, SE3: Inte-
gration, SE4: Feedback analyt-
ics; UX1: Comparison UI, UX2:
Feedback UI, UX3: Feedback ad-
min UI
5 Onboarding,
video, and
garbage solution
7 (Garbage
point), 12
(Video instruc-
tions)
SE1: Garbage backend logic,
SE2: Video integration, SE3:
Garbage solution UI, SE4: Test-
ing; UX1: Video content, UX2:
Garbage UI, UX3: Onboarding
UI
Continued on next page
6.1 EVALUATION USING MYFLAVORIA PROJECT 56
Table 6.2 – continued from previous page
Sprint Sprint Goal Selected User
Stories
Assigned Workers
6 Refine menu fea-
tures and multi-
lingual UI
2 (Menu re-
finements), 4
(Week’s menu
refinements),
13 (Language
finalization)
SE1: Menu logic refinements,
SE2: Language switching, SE3:
Menu UI testing, SE4: Language
testing; UX1: Menu UI polish,
UX2: Language UI review, UX3:
Testing
7 User feedback &
polish
1 (Dietary com-
parison polish),
11 (Feedback
improvements),
Bug fixes
SE1: Dietary improvements, SE2:
Feedback UI/UX, SE3: Bug fix-
ing, SE4: Testing; UX1: UI feed-
back, UX2: Feedback UI, UX3:
Testing
8 Final review,
documentation,
release
All stories (final
testing, docs, re-
lease prep)
SE1: Final testing, SE2: Doc-
umentation, SE3: Release prep,
SE4: Deployment; UX1: User
guide, UX2: Final UI polish,
UX3: Accessibility review
6.1.3 Overall analysis
A detailed analysis is conducted by manually reviewing the historical sprint docu-
mentation of the MyFlavoria project and comparing it to the sprint plans generated
by GenSP. The comparison focused on specific sprint planning criteria, such as com-
pleteness of backlog items, alignment between sprint goals and backlog, workload
distribution among team members, and identification of task dependencies. The
6.1 EVALUATION USING MYFLAVORIA PROJECT 57
objective is to evaluate how effectively GenSP addresses identified challenges in tra-
ditional sprint planning.
• Challenge: Incomplete Backlog Items
Backlog items are sometimes added without clear descriptions or well-defined
acceptance criteria. This lack of detail can cause confusion and lead to mis-
understandings during development and testing.
GenSP Approach: GenSP can break down high-level backlog items into
smaller tasks and suggest relevant details. This provides a starting point for
the team to clarify requirements and acceptance criteria, leading to better
understanding and smoother development.
• Challenge: Misalignment Between Sprint Goals and Backlog
In the initial sprint, the backlog often does not reflect the actual sprint goals.
Both the goals and the user stories may be unclear, and their connection to
each other is often weak. This misalignment can make it difficult for the team
to focus and measure progress.
GenSP Approach: GenSP helps by suggesting backlog items that are closely
linked to the stated sprint goals. This ensures that each sprint has a clear
purpose and that the selected user stories directly contribute to achieving
the sprint objectives. It also encourages meaningful team discussions about
priorities.
• Challenge: Unbalanced Workload Distribution
In some sprints, senior team members are overloaded with work while oth-
ers are underutilized. This creates bottlenecks and can slow down the team’s
overall progress.
GenSP Approach: GenSP considers each team member’s skills and cur-
rent workload when recommending task assignments. This results in a more
6.1 EVALUATION USING MYFLAVORIA PROJECT 58
balanced distribution of work, reduces bottlenecks, and allows everyone to
contribute according to their strengths.
• Challenge: Overlooked Task Dependencies
Dependencies between tasks can be missed during planning, which may cause
delays or require rework later in the sprint. For example, a user interface can-
not be implemented before its design is ready.
GenSP Approach: GenSP automatically identifies and highlights potential
dependencies during the task breakdown phase. This allows the team to ad-
dress dependencies early, reducing the risk of delays and rework during the
sprint.
Disadvantages
Despite its benefits, the GenSP tool also demonstrated certain limitations during the
analysis. Firstly, GenSP occasionally generated unnecessary subtasks that were not
relevant to completing the user story. For example, for the story "As a diner/allergic,
I want to be able to inspect the ingredients of each food option. . . ," the tool suggested
a subtask to "Add ingredient info to DB". However, in the context of the project,
the ingredient information was already managed by an existing API, making this
subtask redundant. This issue could be mitigated by providing clearer user story
descriptions and more precise acceptance criteria.
In a few instances, GenSP appeared to overestimate story points, potentially leading
to inaccurate workload planning. This, too, highlights the importance of well-defined
user stories to support more accurate estimations.
The tool sometimes assigned testing tasks to multiple developers within the same
sprint, when it would have been more efficient to delegate such tasks to a single
developer, freeing others to focus on additional development work.
The evaluation with the MyFlavoria project shows that GenSP improves Agile
6.2 SURVEY RESULTS 59
sprint planning by refining backlogs, aligning backlog items with sprint goals, pro-
viding effort estimates, and balancing task assignments. While GenSP addresses
several challenges of traditional methods and enhances planning clarity, it still has
some shortcomings that require human oversight. Overall, integrating GenAI into
sprint planning can lead to more effective and predictable project outcomes.
6.2 Survey Results
In this section, the survey results with participants’ demographics, quantitative as
well as qualitative results, and implications are discussed. The complete set of survey
questions is provided in the appendix section.
6.2.1 Survey Demographics
A total of nine software engineers and developers voluntarily participated in this
survey. All participants currently hold positions as software engineers or developers
(100%), ensuring that the feedback reflects the perspective of professionals actively
engaged in software development.
Table 6.3: Participant’s current company location
Company Location (Country) Participant Count Percentage
Finland 3 33.3%
United States 2 22.2%
Bangladesh 2 22.2%
Sweden 1 11.1%
United Kingdom 1 11.1%
The respondents represent a diverse set of countries and work environments. As
shown in Table 6.3, participants are employed in Finland (33.3%), the United States
(22.2%), Bangladesh (22.2%), Sweden (11.1%), and the United Kingdom (11.1%).
This international mix brings a range of experiences and viewpoints from Scrum
teams operating in different organizational and cultural settings.
6.2 SURVEY RESULTS 60
The respondents are highly experienced in software development, with the ma-
jority occupying senior roles. Table 6.4 details their years of experience:
Table 6.4: Participant’s software development experience
Experience(Years) Participant Count Percentage
3–5 years 1 11.1%
6–10 years 5 55.6%
More than 10 years 3 33.3%
This shows that almost 90% of the participants have more than six years of in-
dustry experience, ensuring that their insights are grounded in extensive professional
practice.
All respondents reported familiarity with the Scrum framework. Most (88.9%)
identified themselves as having intermediate experience, typically as regular Scrum
team members, while one respondent (11.1%) indicated advanced knowledge, in-
cluding experience as a Scrum Master or Product Owner.
Table 6.5: Familiarity with Scrum
Company Location (Country) Participant Count Percentage
Intermediate (Scrum team member) 8 88.9%
Advanced (Scrum Master/Product Owner) 1 11.1%
Additionally, all participants (100%) reported using project management tools
such as Jira, Trello, or similar software, indicating familiarity with common Agile
project practices.
All survey participants reported familiarity with GenAI tools, including tech-
nologies such as ChatGPT, Gemini, or GitHub Copilot. However, 55.6% indicated
that they had not previously used GenAI specifically for sprint planning, while 44.4%
reported using such tools for sprint planning only occasionally or on a trial basis.
As presented in Table 6.4, the participant group includes software professionals
with experience in Scrum, project management, and exposure to GenAI technologies.
These characteristics provide important context for interpreting their perspectives
6.2 SURVEY RESULTS 61
on the integration of GenAI into Scrum sprint planning. Prior to completing the
survey, participants were shown a live demonstration or video walkthrough of the
GenSP tool to ensure they understood how GenAI can be applied in sprint planning.
6.2.2 Quantitative Results
This section presents the analysis of quantitative data collected through Likert-scale
survey questions. The results are organized according to key use cases of GenAI in
sprint planning, including backlog creation, story point estimation, task breakdown,
and task assignment. Each use case is evaluated based on participants’ responses,
providing measurable insights into the perceived value, usefulness, and potential
challenges of integrating GenAI into the sprint planning process.
GenAI in Sprint Backlog Generation
The results suggest that GenAI is viewed positively for supporting sprint backlog
creation. As shown in Table 6.6, 77.8% of respondents rated GenAI as valuable for
creating the sprint backlog, citing its ability to suggest user stories and potentially
reduce human error and bias. No respondents expressed a negative view, while 22.2
Additionally, the survey asked about GenAI’s ability to identify missing details,
dependencies, or acceptance criteria for items proposed for the sprint backlog. Here,
66.7% of participants saw GenAI as valuable, though a small proportion (22.2%)
expressed skepticism. This finding suggests that while most practitioners appreciate
the AI’s potential for refining backlog items and flagging gaps, some may be cautious
about relying on AI for complex, context-dependent decisions.
When it comes to prioritizing or ordering items within the sprint backlog, re-
sponses were more mixed: 44.4% saw GenAI as helpful, 44.4% were neutral, and
11.1% were negative. This result implies that, while GenAI is appreciated for its
automation and analytical power, practitioners still rely on their experience and
6.2 SURVEY RESULTS 62
Table 6.6: Survey Results on GenAI in Sprint Backlog Generation
Survey Question Positive Neutral Negative
How valuable do you think generative AI
could be in helping to create the sprint backlog
(i.e., suggesting user stories based on sprint
goal, capacity, and previous sprint informa-
tion, reducing human error and bias)?
77.8% 22.2% 0%
How valuable could generative AI be in iden-
tifying missing details, dependencies, or ac-
ceptance criteria for items proposed for the
sprint backlog?
66.7% 11.1% 22.2%
How helpful could generative AI be in sug-
gesting prioritization or ordering of items
within the sprint backlog?
44.4% 44.4% 11.1%
understanding of project priorities for final decision-making.
Further insights were gathered through this question: “In your opinion, during
sprint backlog creation, which area could benefit the most from the use of generative
AI?” the most frequent responses were backlog item refinement improving descrip-
tions and acceptance criteria (77.8%), sprint backlog suggestions based on goal and
capacity (66.7%), and reprioritization of backlog items (33.3%). These results rein-
force the view that GenAI is especially well-suited to tasks involving analysis and
refinement of backlog content, as well as making context-aware suggestions.
However, participants also highlighted key challenges and limitations of using
GenAI in this context. The most common concern was understanding complex
business logic and dependencies (9 votes). This reflects that GenAI may not yet
fully capture the deeper nuances and cross-functional dependencies that experienced
team members can identify.
In summary, the survey results show that experienced Scrum practitioners see
significant value in using GenAI for sprint backlog generation, particularly for back-
log refinement and suggesting backlog items aligned with sprint goals and capacity.
However, they also recognize the current limitations of AI in understanding complex
business logic and dependencies.
6.2 SURVEY RESULTS 63
GenAI in Story Points Estimation
The survey investigated participants’ perceptions of using GenAI for estimating
story points during sprint planning. The results reveal mixed attitudes toward the
reliability and usefulness of AI-generated estimates. When asked whether they agree
that "Generative AI can estimate user story points based on story title, descriptions,
and historical data," only 33.3% responded positively, while 22.2% were neutral,
and a notable 44.4% expressed disagreement. This indicates that, although some
practitioners see value in AI-powered estimation, many remain cautious about fully
trusting these automated outputs.
However, when it comes to the collaborative aspect of estimation, the responses
were more favorable. Over half of the participants (55.6%) agreed that "AI-proposed
estimates or insights help our team reach consensus on story point estimations more
quickly," suggesting that GenAI can serve as a useful facilitator in team discussions.
Similarly, 55.6% said they would be comfortable using an AI-generated estimate
as a starting point for team discussion, rather than relying solely on traditional
methods like Planning Poker. Still, one-third of respondents expressed hesitation or
discomfort with this approach.
Table 6.7: Survey Results on GenAI in Story Points Estimation
Survey Question Positive Neutral Negative
"Generative AI can estimate user story
points based on story title, descriptions and
historical data." How much do you agree with
this statement?
33.3% 22.2% 44.4%
“AI-proposed estimates or insights help our
team reach consensus on story point estima-
tions more quickly.” – Please rate your agree-
ment.
55.6% 11.1% 33.3%
Would you be comfortable using an AI-
generated estimate as a starting point for
team discussion? (e.g., instead of Planning
Poker)?
55.6% 11.1% 33.3%
6.2 SURVEY RESULTS 64
Participants also identified key challenges in leveraging GenAI for story point
estimation. The most frequently mentioned concern was the AI’s ability to handle
novel or complex user stories that require a deep contextual understanding (8 votes).
This feedback underscores the current limitations of GenAI in dealing with the
nuances and complexities that arise in real-world software projects.
Overall, fewer participants are convinced that GenAI can independently provide
accurate story point estimates. However, GenAI can be valuable to speed up team
consensus and as a starting point for discussions.
GenAI in Story/Task Breakdown
The survey explored participants’ perceptions of using GenAI to break down user
stories into tasks or subtasks, as well as its ability to identify dependencies between
those tasks. The findings show that a majority of respondents (66.7%) agree that
GenAI is helpful in creating clear lists of tasks or subtasks from user stories. This in-
dicates a positive attitude toward leveraging AI to support this traditionally manual
and sometimes inconsistent aspect of sprint planning. Only 11.1% of respondents
expressed a negative view, suggesting that most software professionals are open to
at least experimenting with AI-assisted breakdowns.
When asked about GenAI’s usefulness in identifying potential task dependen-
cies—either within a single user story or across multiple stories—over half of the
participants (55.6%) viewed this as a helpful feature, while 44.4% remained neu-
tral, and no participants expressed a negative view. This result highlights that,
while some teams are optimistic about AI’s ability to surface hidden or complex de-
pendencies, others may still be evaluating its reliability and accuracy in real-world
scenarios.
The most common concern, cited by eight participants, was the challenge of
“understanding the specific technical implementation approach.” This underscores
6.2 SURVEY RESULTS 65
Table 6.8: Survey Results on GenAI in Story/Task Breakdown
Survey Question Positive Neutral Negative
“Generative AI is helpful in breaking down
user stories into a clear list of tasks or sub-
tasks.” – How much do you agree?
66.7% 22.1% 11.1%
How helpful could generative AI be in identi-
fying potential task dependencies (within the
story or to other stories)?
55.6% 44.4% 0%
a critical limitation: while GenAI can effectively suggest initial breakdowns and
surface potential dependencies, it may not always capture the complex technical
details or context-specific solutions that teams require for successful delivery.
In summary, the survey results indicate that participants see potential for GenAI
to assist with story and task breakdowns, as well as identifying dependencies.
GenAI in Task Distribution
The survey also examined how practitioners view the use of GenAI in breaking down
user stories into tasks and distributing those tasks among team members. The
responses indicate a generally positive outlook, with 66.7% of participants rating
GenAI as helpful in suggesting task assignments based on individual skills, experi-
ence, or development goals. This suggests that most respondents believe GenAI can
add value by leveraging available data to make better assignment decisions.
In addition, a majority of participants (55.6%) believe that GenAI could be valu-
able in balancing workload across team members during sprint planning. This ability
to recommend fair and efficient task distribution is seen as a potential benefit, espe-
cially for larger teams or projects where manual balancing can be time-consuming
and prone to oversight. However, 33.3% were neutral, and 11.1% viewed this applica-
tion less favorably, reflecting that there are still reservations about the technology’s
readiness and real-world effectiveness.
Participants were also asked to identify the main challenges they see in using
6.2 SURVEY RESULTS 66
Table 6.9: Survey Results on GenAI in Task Distribution
Survey Question Positive Neutral Negative
How helpful could generative AI be in sug-
gesting task assignments based on individual
skills, experience, or development goals?
66.7% 11.1% 22.2%
How valuable could generative AI be in help-
ing to balance the workload across team mem-
bers during sprint planning?
55.6% 33.3% 11.1%
GenAI for task distribution. The top concerns, each cited by six respondents, were
accurately assessing individual skills and handling team preferences in task distri-
bution by GenAI. These concerns highlight the complex, human-centric nature of
task assignment.
In summary, the results suggest that software professionals are optimistic about
the potential of GenAI to support task breakdown and assignment, particularly as
a tool to enhance fairness and efficiency.
Overall Effectiveness of GenAI in Sprint Planning
The survey asked participants whether they believed GenAI could improve the over-
all effectiveness of their sprint planning process. The results show that a major-
ity—55.6%—agreed that GenAI offers significant value, while 22.2% were neutral
and 22.2% disagreed. This distribution suggests that most experienced software
developers and Scrum practitioners see clear potential in GenAI.
To better understand where GenAI adds the most value, respondents were also
asked which aspects of sprint planning benefit most from AI support (Figure 6.1).
The most commonly selected area was improving planning efficiency, which is cited
by seven participants. This aspect highlights GenAI’s ability to save time and
reduce manual effort in preparing for sprints. User story refinement was the next
most frequent response (six votes), which indicates that many teams recognize the
usefulness of AI in clarifying requirements and acceptance criteria. Sprint backlog
6.2 SURVEY RESULTS 67
creation and task breakdown/planning each received four votes, suggesting that
GenAI is also seen as a valuable assistant in building and structuring the work for
each sprint.
Figure 6.1: GenAI Benefits in Sprint Planning (Survey Results)
Despite some reservations, the survey results indicate that many participants rec-
ognize clear value in GenAI tools. Most participants agree that GenAI can positively
impact sprint planning, especially for initial backlog creation and task breakdown.
Respondents view GenAI not only as a means of automating routine tasks but also
as a facilitator that can support more thoughtful and effective planning discussions.
These findings suggest that GenAI tools have the potential to enhance both the
efficiency and quality of sprint planning overall.
6.2.3 Qualitative Results
To complement the quantitative analysis, participants were invited to share their
open-ended thoughts on the use of GenAI in sprint planning in the survey. The
collected responses provide nuanced perspectives on the potential benefits, current
limitations, and considerations for the adoption of GenAI in Agile processes.
6.2 SURVEY RESULTS 68
Benefits and Opportunities
Several respondents highlighted clear advantages in using GenAI as a supportive
tool during sprint planning. Participants noted that GenAI could significantly save
time by automating routine tasks and providing smart suggestions for user stories,
task breakdown, and estimation. One developer remarked,
"Using GenAI in sprint planning saves time, gives smart suggestions,
improves task clarity, and helps teams plan better together."
One participant noted that GenAI can be very helpful in sprint planning, par-
ticularly for drafting user stories, distributing tasks among team members, and
organizing work.
"I think using generative AI in sprint planning can be really helpful, espe-
cially for things like drafting user stories, helping distribute tasks among
team members, and organizing tasks. It saves time and can surface useful
suggestions or insights we might overlook."
This statement was echoed by others, who saw GenAI as a means to generate
draft user stories, identify dependencies, and offer task distribution suggestions, thus
streamlining the planning process. The following two statements from participants
highlight the benefits and opportunities of using GenAI in sprint planning.
"It can be a purposeful tool for prioritizing tasks."
"I didn’t used GenAI for sprint planning yet. But I think GenAI will be
great tool to write use stories and task break down. If privacy ensured
then I think GenAI can be helpful to do task break down, distribution and
estimation by feeding previos sprints data."
6.2 SURVEY RESULTS 69
One participant suggested that organizations could first try using GenAI tools
with small teams or on smaller projects. This approach would allow them to see
how well the tools work before using them on larger projects.
"Good idea. Small team and project can do trial and see results how
effective it was and based on large scale projects can try to adapt it."
Concerns and Limitations
While many participants recognized the potential of GenAI in sprint planning, some
expressed concerns about privacy, security, and the sensitivity of data. These respon-
dents felt that the use of GenAI should be carefully controlled to protect sensitive
information. For example, one participant remarked:
"I think we could use GenAI in a controlled manner considering secu-
rity and sensitive information that should not be exposed. Additionally,
we need to improve GenAI models to make them more accurate in the
future."
A few participants also noted that current GenAI models may not yet be mature
enough for reliable use in sprint planning. They pointed out that these tools are
not yet able to understand or address the finer details that are often involved in the
planning process. As one respondent stated:
"GenAI hasn’t reached to a level of assessing the fine details required in
sprint planning."
Despite these limitations, there was optimism about the future of GenAI in this
area. Some participants expressed hope that ongoing improvements would eventu-
ally make GenAI a valuable resource for sprint planning and task distribution. As
one participant commented:
6.3 MAPPING RESEARCH QUESTIONS WITH RESULTS 70
"Still a long way to go for generative AI to be used in professional sprint
planning. But I am sure some day AI will help tremendously in sprint
planning and task distribution."
Balancing GenAI with Human Judgment
While recognizing the promise of GenAI, multiple respondents emphasize that hu-
man judgment and team context remain essential for effective sprint planning. As
one participant states:
"I don’t think it should replace human judgment since every team has its
own way of working, and AI doesn’t always understand the full context.
I see it more as a supportive tool that helps streamline the process, not
make the decisions for us."
Overall, the qualitative feedback indicates that experienced software practition-
ers view GenAI as a promising assistant for sprint planning—capable of improving
efficiency, generating useful suggestions, and supporting the planning process. How-
ever, its current limitations, the irreplaceable value of human expertise, and impor-
tant concerns around privacy and context highlight the need for careful, incremental
adoption and ongoing improvements in GenAI technologies.
6.3 Mapping Research Questions with Results
The research questions guide both the practical evaluation of the GenSP tool and
the structured survey of Agile practitioners. By mapping the research questions to
the outcomes of these two approaches, the effectiveness and limitations of GenAI in
sprint planning can be thoroughly assessed.
RQ1 addresses the value and effectiveness of GenAI in sprint backlog creation;
both GenSP’s practical outputs and survey responses provide evidence of its bene-
6.3 MAPPING RESEARCH QUESTIONS WITH RESULTS 71
fits. GenSP was able to suggest relevant user stories that aligned well with sprint
goals and helped identify missing details or dependencies that are often overlooked
during manual planning. Survey participants also acknowledged that GenAI sup-
port in backlog creation improved the refinement process and reduced human error,
although some noted challenges in handling complex business logic.
RQ2 focuses on GenAI’s performance in user story point estimation. The GenSP
tool demonstrated that GenAI could generate story point estimates with clear ra-
tionale, which helped facilitate team discussion and reduce subjective bias. Survey
results showed a split in confidence, with some participants finding AI-generated
estimates useful as a starting point, while others expressed concerns about the AI’s
ability to handle novel or complex stories. Nevertheless, the majority agreed that
GenAI estimates could help teams reach consensus faster compared to purely manual
methods.
RQ3 explores how GenAI assists in breaking down user stories into actionable
tasks and identifying dependencies. GenSP’s task breakdowns improve clarity and
help teams recognize dependencies early. Survey respondents agreed that GenAI
could generate detailed task lists and highlight potential dependencies, although
there was some hesitation about relying solely on AI for technical implementation
details. Many saw AI-driven breakdowns as a valuable draft for team refinement.
RQ4 examines GenAI’s role in task assignment and workload balancing. The
GenSP tool’s approach to assigning tasks results in more balanced workloads, as seen
in the comparison with MyFlavoria’s sprint history. Survey responses support this
finding, with participants noting that GenAI could assist in distributing work fairly,
but they also highlighted challenges in assessing individual skills and preferences
purely through AI.
Finally, RQ5 considers the overall impact of GenAI on the effectiveness, effi-
ciency, and quality of sprint planning. Both the GenSP evaluation and the survey
6.4 DISCUSSION 72
indicated that GenAI improves sprint planning, enhances backlog and task clarity,
and supports better workload management. The majority of survey participants
agreed that GenAI made the sprint planning process more effective, especially in
saving time and reducing manual effort. However, the participants emphasize the
continued need for human oversight.
In summary, by mapping each research question to the findings from GenSP’s
tool-based evaluation and the practitioner survey, this study demonstrates that
GenAI holds significant promise for addressing many of the persistent challenges
in Scrum sprint planning. This also highlights the areas where further development
in GenAI models is needed for performing more complex scenarios in sprint planning.
6.4 Discussion
The findings of this thesis offer valuable insights into the practical application and
impact of GenAI in sprint planning. This section discusses these results in detail.
6.4.1 Evaluating GenSP Tool Data
When comparing GenSP-generated sprint plans with the MyFlavoria project’s his-
torical sprint data, several advantages become evident. The GenAI-assisted plans
are more closely aligned with sprint goals. The tool also demonstrates improved
workload balancing and prompted clearer backlog definitions. GenSP’s structured
approach to selecting user stories, estimating effort, and assigning tasks reduced the
likelihood of overcommitment and made hidden dependencies more visible during
planning. This tool lays a foundation for the feasibility of using GenAI directly in
the sprint planning process.
6.4 DISCUSSION 73
6.4.2 Interpretation of Survey Results
The survey results indicate a generally positive attitude among experienced software
developers toward using GenAI in sprint planning. The majority of participants
recognize GenAI’s value in automating backlog refinement, generating user stories,
and supporting task breakdown. In particular, respondents highlighted time savings,
reduced manual effort, and the ability to surface suggestions or dependencies that
might otherwise be overlooked. In summary, the experienced scrum practitioners
mostly support the usefulness of using GenAI in sprint planning.
However, the results also reveal a note of caution. While most participants
would be comfortable using AI-generated sprint backlogs and task breakdowns as a
starting point, there remains some skepticism about relying solely on AI for complex
estimations or decisions that require deep contextual knowledge.
6.4.3 Key Contribution
This research makes certain key contributions to the study of GenAI in sprint plan-
ning. First, it offers a comprehensive review of the current challenges in traditional
sprint planning. This thesis also highlights the gaps in existing research regarding
the integration of GenAI into this process.
A key practical contribution is the design and development of GenSP. The tool
provides ChatGPT-supported sprint backlog generation, story point estimation, task
breakdown, and task distribution.
Additionally, the thesis contributes empirical insights by a structured survey of
experienced Scrum practitioners. The survey results, along with the qualitative
and quantitative analyses, provide evidence of the perceived value, limitations, and
potential of GenAI in supporting sprint planning.
6.4 DISCUSSION 74
6.4.4 Limitations
There are several limitations in the thesis that should be acknowledged. First, the
evaluation of the GenSP tool is not conducted with a real Scrum team in a live
development environment. Instead, the tool is tested using the backlog and sprint
data of an existing software project. Using a real project would show more closely
how GenAI supports the sprint planning process.
Second, the number of survey respondents was limited. Although all participants
have prior experience in software development and Agile methodologies, they did
not use the GenSP tool directly. As a result, their responses reflect expert opinions
based on their knowledge and understanding, rather than hands-on experience with
the tool. While this helped ensure relevance, it may also have introduced some bias
and limited the diversity of perspectives represented in the results. Larger-scale
studies involving a broader range of teams and organizations would strengthen the
generalizability of the findings.
Lastly, GenSP depends on the OpenAI GPT API, so its outputs are subject
to the limitations of LLMs, such as hallucination [43], limited domain knowledge,
or context misinterpretation, despite efforts to improve reliability through prompt
engineering.
Despite these limitations, the thesis lays a foundation for further exploration and
highlights important considerations for both researchers and practitioners interested
in integrating GenAI into Agile project management.
7 Conclusion
This thesis explored how Generative AI (GenAI) can support and improve sprint
planning in Agile software development. Using a literature review, prototype tool
development, and a survey of Agile practitioners, the research found that GenAI
can aid the sprint planning process.
The GenSP tool, created for this research, showed that GenAI can automate key
tasks such as generating sprint backlogs based on team goals and capacity, estimating
story points, breaking down user stories into actionable tasks, and assigning work
more evenly. These features were tested using real project data and compared to
traditional planning methods.
Survey responses from experienced Agile practitioners highlighted the practical
value of GenAI. Most agreed that it can save time, provide helpful suggestions, and
improve task clarity. However, they also noted challenges, such as understanding
complex requirements, assessing individual skills, and ensuring security and pri-
vacy. Importantly, respondents emphasized that GenAI should support—not re-
place—human decision-making.
The results suggest that GenAI can make sprint planning more efficient and con-
sistent, reducing manual effort and facilitating better team discussions. Nonetheless,
human oversight remains essential for interpreting AI recommendations and main-
taining alignment with team objectives.
7.1 FUTURE WORK 76
7.1 Future Work
Although this thesis demonstrated the potential of GenAI in supporting Agile sprint
planning, several avenues remain open for further exploration and improvement.
First, a key next step is to test GenAI-powered tools like GenSP in live, ongoing
projects managed with Scrum. Evaluating the tool during active sprints would
provide richer insights into its practical value, user acceptance, and impact on team
performance in real-world settings. In addition, conducting surveys among a wider
group of practitioners from diverse organizations and backgrounds would further
strengthen the understanding of GenAI’s applicability and effectiveness in various
Agile environments.
Second, integrating GenSP into the Sprint Retrospective process could offer a
feedback loop that continually improves AI-generated outputs. By allowing teams
to review, discuss, and rate the accuracy and usefulness of GenAI suggestions during
retrospectives, the system can adapt to each team’s unique context and learn from
previous sprints.
Finally, future research should compare different LLMs to identify the most
effective one for Agile sprint planning. Assessing models based on accuracy, response
quality, explainability, and integration ease will help determine which LLMs best
meet the needs of Scrum teams.
Exploring these directions will strengthen the practical application of GenAI
in Agile environments and contribute to the development of more adaptive and
intelligent project management solutions that support Agile practices.
References
[1] M. U. Haque, I. Dharmadasa, Z. T. Sworna, R. N. Rajapakse, and H. Ahmad,
“"i think this is the most disruptive technology": Exploring sentiments of chat-
gpt early adopters using twitter data”, arXiv preprint arXiv:2212.05856, 2022.
[2] V. Corso, L. Mariani, D. Micucci, and O. Riganelli, “Assessing ai-based code
assistants in method generation tasks”, in Proceedings of the 2024 IEEE/ACM
46th International Conference on Software Engineering: Companion Proceed-
ings, 2024, pp. 380–381.
[3] B. Sherifi, K. Slhoub, and F. Nembhard, “The potential of llms in automating
software testing: From generation to reporting”, arXiv preprint arXiv:2501.00217,
2024.
[4] C. Ebert and P. Louridas, “Generative ai for software practitioners”, IEEE
Software, vol. 40, no. 4, pp. 30–38, 2023.
[5] H. K. Flora and S. V. Chande, “A systematic study on agile software de-
velopment methodologies and practices”, International Journal of Computer
Science and Information Technologies, vol. 5, no. 3, pp. 3626–3637, 2014.
[6] M. Hron and N. Obwegeser, “Why and how is scrum being adapted in practice:
A systematic review”, Journal of Systems and Software, vol. 183, p. 111 110,
2022.
REFERENCES 78
[7] Scrum.org, What is scrum?, https://www.scrum.org/learning-series/
what-is-scrum, Accessed: 2024-06-27, 2020.
[8] S. Alliance, “Sprint planning meeting: Why it’s critical for agile success”,
Scrum Alliance Resource Library, 2025, Reviewed by: Bernie Maloney, Mad-
hur Kathuria, and Raúl Herranz. [Online]. Available: https://resources.
scrumalliance.org/Article/sprint-planning-meeting.
[9] K. Schwaber and J. Sutherland, The scrum guide, Accessed: 2024-06-27, 2020.
[Online]. Available: https://scrumguides.org/scrum-guide.html.
[10] K. Melnyk, V. Hlushko, and N. Borysova, “Decision support technology for
sprint planning”, Radio Electronics, Computer Science, Control, no. 1, pp. 135–
145, 2020.
[11] M. Golfarelli, S. Rizzi, and E. Turricchia, “Multi-sprint planning and smooth
replanning: An optimization model”, Journal of systems and software, vol. 86,
no. 9, pp. 2357–2370, 2013.
[12] M. Žáček, A. Hamplová, J. Tyrychtr, and I. Vrana, “Improvements for the
planning process in the scrum method”, Applied Sciences, vol. 15, no. 1, p. 202,
2024.
[13] L. Alsaber, E. Al Elsheikh, S. Aljumah, and N. S. M. Jamail, ““scrumbear”
framework for solving traditional scrum model problems”, Bulletin of Electrical
Engineering and Informatics, vol. 10, no. 1, pp. 319–326, 2021.
[14] A. R. Hevner, S. T. March, J. Park, and S. Ram, “Design science in information
systems research”, MIS quarterly, pp. 75–105, 2004.
[15] M. Kasunic, Designing an effective survey, 2005.
[16] B. Wang et al., “Towards understanding chain-of-thought prompting: An em-
pirical study of what matters”, arXiv preprint arXiv:2212.10001, 2022.
REFERENCES 79
[17] E. Guglielmi, V. Arnoudova, G. Bavota, R. Oliveto, and S. Scalabrino, “How
do copilot suggestions impact developers’ frustration and productivity?”, arXiv
preprint arXiv:2504.06808, 2025.
[18] C. Stryker and E. Kavlakoglu, What is artificial intelligence (ai)?, Published
9 August 2024; accessed 2025-06-14, 2024. [Online]. Available: https://www.
ibm.com/think/topics/artificial-intelligence.
[19] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 3rd.
Upper Saddle River, NJ: Prentice Hall, 2010, isbn: 978-0-13-604259-4.
[20] M. E. Klontzas, S. C. Fanni, and E. Neri, Eds., Introduction to Artificial Intelli-
gence (Imaging Informatics for Healthcare Professionals). Springer, 2023, isbn:
978-3-031-25927-2. [Online]. Available: https://doi.org/10.1007/978-3-
031-25928-9.
[21] S. M. Mian, M. S. Khan, M. Shawez, and A. Kaur, “Artificial intelligence
(ai), machine learning (ml) & deep learning (dl): A comprehensive overview
on techniques, applications and research directions”, in 2024 2nd International
Conference on Sustainable Computing and Smart Systems (ICSCSS), IEEE,
2024, pp. 1404–1409.
[22] A. Dubey, “Usage of deep learning in recent applications”, Archives of Mate-
rials Science and Engineering, vol. 115, no. 2, 2022.
[23] S. S. Sengar, A. B. Hasan, S. Kumar, and F. Carroll, “Generative artificial
intelligence: A systematic review and applications”, Multimedia Tools and Ap-
plications, pp. 1–40, 2024.
[24] Z. Sordo, E. Chagnon, and D. Ushizima, “A review on generative ai for text-
to-image and image-to-image generation and implications to scientific images”,
arXiv preprint arXiv:2502.21151, 2025.
REFERENCES 80
[25] S. Feuerriegel, J. Hartmann, C. Janiesch, and P. Zschech, “Generative ai”,
Business & Information Systems Engineering, vol. 66, no. 1, pp. 111–126,
2024.
[26] L. Banh and G. Strobel, “Generative artificial intelligence”, Electronic Markets,
vol. 33, no. 1, p. 63, 2023.
[27] R. K. Gatla, A. Gatla, P. Sridhar, D. G. Kumar, and D. N. M. Rao, “Advance-
ments in generative ai: Exploring fundamentals and evolution”, in 2024 Inter-
national Conference on Electronics, Computing, Communication and Control
Technology (ICECCC), IEEE, 2024, pp. 1–5.
[28] Stack Overflow. “Stack overflow developer survey 2024: AI developer tools”.
Accessed: 2024-06-27, Stack Overflow, Accessed: Jun. 27, 2024. [Online]. Avail-
able: https://survey.stackoverflow.co/2024/ai#developer-tools.
[29] S. Peng, E. Kalliamvakou, P. Cihon, and M. Demirer, “The impact of ai on de-
veloper productivity: Evidence from github copilot”, arXiv preprint arXiv:2302.06590,
2023.
[30] Z. Zheng, K.-Y. Chen, X.-Y. Cao, X.-Z. Lu, and J.-R. Lin, “Llm-funcmapper:
Function identification for interpreting complex clauses in building codes via
llm”, arXiv preprint arXiv:2308.08728, 2023.
[31] G. Bandarupalli, “Code reborn ai-driven legacy systems modernization from
cobol to java”, arXiv preprint arXiv:2504.11335, 2025.
[32] Y. Fu et al., “Creativity in the age of ai: Evaluating the impact of gener-
ative ai on design outputs and designers’ creative thinking”, arXiv preprint
arXiv:2411.00168, 2024.
[33] S. Dandotiya, “Generative ai for software testing: Harnessing large language
models for automated and intelligent quality assurance [j]”, International Jour-
nal of Science and Research Archive, vol. 14, no. 1, pp. 1931–1935, 2025.
REFERENCES 81
[34] C. Arora, J. Grundy, and M. Abdelrazek, “Advancing requirements engineer-
ing through generative ai: Assessing the role of llms”, in Generative AI for
Effective Software Development, Springer, 2024, pp. 129–148.
[35] T. Rahman and Y. Zhu, “Automated user story generation with test case
specification using large language model”, arXiv preprint arXiv:2404.01558,
2024.
[36] T. Dyba and T. Dingsoyr, “What do we know about agile software develop-
ment?”, IEEE software, vol. 26, no. 5, pp. 6–9, 2009.
[37] A. Alliance, Principles behind the agile manifesto, https://agilemanifesto.
org/principles.html, 2001.
[38] M. Coram and S. Bohner, “The impact of agile methods on software project
management”, in 12th IEEE International Conference and Workshops on the
Engineering of Computer-Based Systems (ECBS’05), IEEE, 2005, pp. 363–
370.
[39] N. Damij and T. Damij, “An approach to optimizing kanban board workflow
and shortening the project management plan”, IEEE transactions on engineer-
ing management, vol. 71, pp. 13 266–13 273, 2021.
[40] A. Putta, M. Paasivaara, and C. Lassenius, “Benefits and challenges of adopt-
ing the scaled agile framework (safe): Preliminary results from a multivocal”, in
International Conference on Product-Focused Software Process Improvement,
Vol. 1, Springer International Publishing, 2018.
[41] M. Hron and N. Obwegeser, “Scrum in practice: An overview of scrum adap-
tations”, in Hawaii International Conference on System Sciences, Curran As-
sociates, Inc., 2018, pp. 4496–4505.
REFERENCES 82
[42] M. A. Beltran, M. I. Ruiz Mondragon, and S. H. Han, “Comparative analysis
of generative ai risks in the public sector”, in Proceedings of the 25th Annual
International Conference on Digital Government Research, 2024, pp. 610–617.
[43] M. T. Hicks, J. Humphries, and J. Slater, “Chatgpt is bullshit”, Ethics and
Information Technology, vol. 26, no. 2, pp. 1–10, 2024.
[44] B. Cabrero-Daniel, “How reliance on genai might limit human creativity and
critical thinking in requirements engineering”, 2025.
[45] M. Abbas, F. A. Jam, and T. I. Khan, “Is it harmful or helpful? examining
the causes and consequences of generative ai usage among university students”,
International Journal of Educational Technology in Higher Education, vol. 21,
no. 1, p. 10, 2024.
[46] Scrum.org. “What is sprint planning?”, Scrum.org. [Online]. Available: https:
//www.scrum.org/learning-series/what-is-scrum/the-scrum-events/
what-is-sprint-planning.
[47] Atlassian. “How to plan a sprint”. Accessed: 2024-06-27. [Online]. Available:
https://www.atlassian.com/agile/scrum/sprint-planning.
[48] R. K. Mallidi and M. Sharma, “Study on agile story point estimation tech-
niques and challenges”, Int. J. Comput. Appl, vol. 174, no. 13, pp. 9–14, 2021.
[49] D. Özkan and A. Mishra, “Agile project management tools: A brief comprative
view”, Cybernetics and Information Technologies, vol. 19, no. 4, pp. 17–25,
2019.
[50] M. O. Zander and M. Meboldt, “Navigating the unknown: Introduction of
an analysis metric for the sprint plan and sprint review in educational agile
development projects”, in Proceedings of the 52nd Annual Conference of SEFI,
Lausanne, Switzerland., Zenodo, 2024.
REFERENCES 83
[51] L. A. Garcia, E. OliveiraJr, and M. Morandini, “Tailoring the scrum framework
for software development: Literature mapping and feature-based support”, In-
formation and Software Technology, vol. 146, p. 106 814, 2022.
[52] S. Yermolaieva, “Communication challenges in agile teams from the commu-
nication theory prospective”, in Proceedings of the 2020 European Symposium
on Software Engineering, 2020, pp. 88–95.
[53] H. K. Dam, T. Tran, J. Grundy, A. Ghose, and Y. Kamei, “Towards effective
ai-powered agile project management”, in 2019 IEEE/ACM 41st international
conference on software engineering: new ideas and emerging results (ICSE-
NIER), IEEE, 2019, pp. 41–44.
[54] A. Bahi, J. GHARI, and Y. Gahi, “Integrating generative ai for advancing
agile software development and mitigating project management challenges.”,
International Journal of Advanced Computer Science & Applications, vol. 15,
no. 3, 2024.
[55] S. Ostrowski, “Using genai in it project management: Case studies, insights and
challenges.”, Scientific Papers of Silesian University of Technology. Organiza-
tion & Management/Zeszyty Naukowe Politechniki Slaskiej. Seria Organizacji
i Zarzadzanie, no. 215, 2025.
[56] M. Choetkiertikul, H. K. Dam, T. Tran, T. Pham, A. Ghose, and T. Menzies,
“A deep learning model for estimating story points”, IEEE Transactions on
Software Engineering, vol. 45, no. 7, pp. 637–656, 2018.
[57] R. K. Mallidi, M. Sharma, and Y. P. Paladugu, “Story point estimate model:
Project development using generative ai (genai) tools”, International Journal
of Computer Applications, vol. 975, p. 8887, 2024.
[58] M. R. Islam and P. Sandborn, “Multimodal generative ai for story point esti-
mation in software development”, arXiv preprint arXiv:2505.16290, 2025.
REFERENCES 84
[59] T. K. Chiu, “The impact of generative ai (genai) on practices, policies and
research direction in education: A case of chatgpt and midjourney”, Interactive
Learning Environments, vol. 32, no. 10, pp. 6187–6203, 2024.
[60] Myflavoria application, https://myflavoria.fi/app/, Accessed: 22 July
2025, 2025.
[61] F. K. Willits, G. L. Theodori, and A. Luloff, “Another look at likert scales”,
Journal of rural social sciences, vol. 31, no. 3, p. 6, 2016.
[62] Amazon Web Services, What is a RESTful API?, https://aws.amazon.com/
what-is/restful-api/, Accessed on July 04, 2025.
[63] Express.js, Express.js 5.x API Reference, https://expressjs.com/en/5x/
api.html, Accessed on July 04, 2025.
[64] MongoDB Inc., MongoDB Documentation, https : / / www . mongodb . com /
docs/, Accessed on July 04, 2025.
[65] MongooseJS, Mongoose Documentation, https://mongoosejs.com/docs/
index.html, Accessed on July 04, 2025.
[66] OpenAI,OpenAI Node.js Library Documentation, https://platform.openai.
com/docs/libraries/node-js-library, Accessed on July 04, 2025.
[67] M. Principe and D. Yoon, “A web application using mvc framework”, in Pro-
ceedings of the International Conference on e-Learning, e-Business, Enterprise
Information Systems, and e-Government (EEE), The Steering Committee of
The World Congress in Computer Science, Computer . . ., 2015, p. 10.
[68] OpenAI. “Api reference: Chat”, OpenAI, Accessed: Jul. 25, 2025. [Online].
Available: https://platform.openai.com/docs/api-reference/chat.
[69] P. Bansal, “Prompt engineering importance and applicability with generative
ai”, Journal of Computer and Communications, vol. 12, no. 10, pp. 14–23,
2024.
Appendix A Survey Questions
The survey questions are available here: https://forms.gle/sE6fubxgCpDKCVtY9
A.1 Generative AI in Sprint Planning: Survey Ques-
tions
1. Contact Email
2. Country of Residence
3. Your Current Role
• Software Engineer/Developer
• DevOps Engineer
• Quality Assurance Engineer (QA)
• Scrum Master
• Product Owner
• Other
4. How many years of professional software development experience do
you have?
• 0–2 years
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-2
• 3–5 years
• 6–10 years
• More than 10 years
5. What is your familiarity with Agile/Scrum methodologies?
• Not familiar (no direct experience)
• Beginner (some exposure)
• Intermediate (regular team member)
• Advanced (Scrum Master/Product Owner)
6. Do you or have you ever used project management tools? (e.g., Jira,
Trello, etc.)
• Yes
• No
7. How frequently do you use generative AI tools (e.g., ChatGPT,
Gemini, Copilot) in your work?
• Frequently (on most days or in most projects)
• Occasionally (a few times a month)
• Rarely (a few times total)
• Never
8. Have you ever used generative AI as part of your Sprint Planning
process?
• Yes, regularly
• Yes, occasionally
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-3
• No, never
9. Which sprint planning activities have you used generative AI for?
• Writing or refining user story descriptions
• Estimating story points or effort for user stories
• Breaking down user stories into tasks/sub-tasks
• Prioritizing or grooming the product backlog
• Drafting or clarifying sprint goals
• I have not used GenAI in any sprint planning activities
• Other
10. How valuable do you think generative AI could be in helping to
create the sprint backlog (suggesting user stories based on sprint
goal, capacity, etc.)? (1 = Not Valuable, 5 = Extremely Valuable)
11. How valuable could generative AI be in identifying missing details,
dependencies, or acceptance criteria for sprint backlog items? (1 =
Not Valuable, 5 = Extremely Valuable)
12. How helpful could generative AI be in suggesting prioritization or
ordering of items within the sprint backlog? (1 = Not Helpful, 5 = Very
Helpful)
13. In your opinion, during sprint backlog creation, which area could
benefit the most from generative AI?
• Sprint backlog item suggestions
• Backlog item refinement
• Reprioritizing backlog items
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-4
• Other
14. What do you see as the biggest challenge in using generative AI for
sprint backlog creation?
• Ensuring sprint backlog alignment with sprint goal
• Quality/Relevance of AI suggestions
• Understanding complex business logic/dependencies
• Other
15. Your Thoughts on Applying GenAI to Sprint Backlog Creation (Open
comment)
16. "Generative AI can estimate user story points based on story title,
descriptions, and historical data." How much do you agree? (1 =
Strongly Disagree, 5 = Strongly Agree)
17. “AI-proposed estimates or insights help our team reach consensus
on story point estimations more quickly.” Rate your agreement. (1
= Strongly Disagree, 5 = Strongly Agree)
18. Would you be comfortable using an AI-generated estimate as a start-
ing point for team discussion? (1 = Not Comfortable, 5 = Very Comfort-
able)
19. What do you see as the biggest challenge in using generative AI for
story point estimation?
• Quality/Accuracy of estimates
• Handling novel or complex stories
• Other
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-5
20. Your Thoughts on Using GenAI for Estimating Story Points in
Sprint Planning (Open comment)
21. “Generative AI is helpful in breaking down user stories into a clear
list of tasks or sub-tasks.” How much do you agree? (1 = Strongly
Disagree, 5 = Strongly Agree)
22. How helpful could generative AI be in identifying potential task
dependencies (within the story or to other stories)? (1 = Not Helpful,
5 = Very Helpful)
23. Would you be comfortable using a generative AI-generated task
breakdown as a starting point for the team to refine? (1 = Not Com-
fortable, 5 = Very Comfortable)
24. What do you see as the biggest challenge in using generative AI for
story-to-task breakdown?
• Generating overly generic or irrelevant tasks
• Understanding the technical implementation approach
• Lack of technical detail in the user story
• Other
25. Your Thoughts on Using GenAI for Breaking Down User Stories
into Tasks (Open comment)
26. How helpful could generative AI be in suggesting task assignments
based on individual skills, experience, or development goals? (1 =
Not Helpful, 5 = Very Helpful)
27. How valuable could generative AI be in helping to balance the work-
load across team members during sprint planning? (1 = Not Valuable,
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-6
5 = Extremely Valuable)
28. What do you see as the biggest challenge in using generative AI for
task distribution?
• Accurately assessing individual skills/capabilities
• Handling team preferences in task distribution
• Perceived fairness/bias in suggestions
• Other
29. Your Thoughts on Using GenAI for Task Distribution in Sprint Plan-
ning (Open comment)
30. “Generative AI, on the whole, can improve the effectiveness of our
sprint planning process.” Please indicate your agreement. (1 = Strongly
Disagree, 5 = Strongly Agree)
31. In your opinion, which aspect of sprint planning benefits the most
from generative AI support?
• User story refinement
• Sprint backlog creation
• Story point estimation
• Task breakdown and planning
• Backlog prioritization
• Improving planning efficiency
• Other
32. What challenges or concerns do you have (or do you foresee) with
using generative AI in sprint planning?
A.1 GENERATIVE AI IN SPRINT PLANNING: SURVEY QUESTIONS A-7
• AI suggestions can be inaccurate
• AI might reduce team discussion/human insight
• Not convinced AI adds value
• Difficulty integrating with existing tools
• Privacy and security concerns
• Other
33. Your Thoughts on Using GenAI in Sprint Planning Overall (Open
comment)