Empowering Business Transformation: The Positive Impact and Ethical Considerations of Generative AI in Software Product Management -- A Systematic Literature Review

1. Introduction

• Generative AI creates novel content such as text, images, or music using statistical learning and approaches including GANs, VAEs, and transformers.Source: §1 · p. 2
• Software product management covers creating, delivering, and maintaining software products, aligning business objectives with customer needs across the product lifecycle.Source: §1 · p. 2

Problem Statement

• The review aims to address limited awareness of GenAI’s potential in software product management and examine its applications and ethical considerations across product-manager activities.Source: §1 · p. 2, Problem Statement · p. 3

2. Significance of the Study

• The study identifies a lack of comprehensive research on generative AI’s applications, impact, and ethical implications in software product management.Source: §2 · p. 3
• The study’s findings offer a holistic understanding of GenAI’s role and value, including potential gains in labor productivity, personalized customer experience, accelerated R&D, and new business models.Source: §2 · p. 3
• The research discusses ethical implications and responsible-use guidelines, aiming to connect academic research with practical application and inform AI strategies and policy discussion.Source: §2 · p. 3

3.1.1 Idea Generation; 3.1.2 Market Research

• Research describes generative AI for idea generation and market research, including GPT-NEO-1.3B’s more coherent linked ideas and GPT’s realistic customer-preference responses.The cited studies also note that language models can hallucinate and return incorrect information.Source: §3.1.1 · p. 4, §3.1.2 · p. 4

3.1.3 Customer Insights

• A conversational AI assistant increased productivity, customer sentiment, and reduced turnover across 5,179 support agents, with the strongest benefits for newer and less-skilled workers.Conversation analysis suggested recommendations helped less-skilled agents communicate more like high-skilled agents.Source: §3.1.3 · p. 5

3.1.4 Product Requirements Engineering

• Malik et al. (2022) reported F1-score improvements of 4% on OpenCoss and 5% on IBM-UAV for their two-phase requirements-conflict detection process.The process uses transformer sentence embeddings and cosine similarity to flag potential conflicts, then named-entity analysis to determine the final set.Source: §3.1.4 · p. 5

3.1.5 Agile Software Development

• GPT2SP outperformed within-project estimates by 34%-57%, cross-project estimates by 39%-49%, and Deep-SE by 6%-47%.It was tested on over 23,000 issues across 16 open-source projects; practitioners valued explanations and found estimation challenging.Source: §3.1.5 · p. 5

3.1.6 Automated Code Generation

• Copilot users completed tasks 55.8% faster, while Codex achieved an average BLEU score of 20.63 for documentation generation under basic settings.The ALSI-Transformer was also developed for code-comment generation, aggregating multimodal information through a Gate Network.Source: §3.1.6 · p. 6

3.1.7 UI/UX Design

• Generative AI may support UX modernization, a complex, labor-intensive process with many manual tasks, although the cited work presents it as a possible solution rather than an established one.The modernization projects involved project managers, UX designers, and software engineers; UX was often less addressed than core code transformation.Source: §3.1.7 · p. 6

3.2 Relevant Studies in Ethics and Privacy

• Ethical and privacy concerns include hallucinations requiring human review, dependence on quality training data, opacity, bias, misuse, and unresolved intellectual-property and copyright questions (Cao et al., 2023).The cited work also raises confidentiality risks from entering confidential information into AI prompts.Source: §3.2 · pp. 6–7

3.3 Relevant Studies in Business and Management

• Business-management studies describe generative AI as a possible aid to structured decision-making and personalized customer service, while noting that customer engagement, loyalty, and new value creation need further research.The research also reports that age and gender influence adoption and use in decision-making.Source: §3.3 · p. 7

4. Review Questions

• The review applies the population, intervention, comparison, outcome, and context framework to product managers and business leaders in software organizations, with no comparison included.The outcome is generative AI applications and ethical considerations in software product management; the questions are presented in Table 1.Source: §4 · p. 8, Table 1

Research Questions and Motivation

• One research question asks how generative AI is applied in software product management and aims to inform product managers considering its use to improve products and services.The review also seeks to identify areas for further research and development.Source: Research Questions and Motivation · p. 8
• The second question asks what ethical challenges generative AI presents and what mitigation strategies are available, addressing concerns about bias, transparency, privacy, and misuse.The motivation calls for examining benefits alongside ethical difficulties so businesses can pursue responsible use.Source: Research Questions and Motivation · p. 8

5. Review Methods

• The qualitative, descriptive review uses current and emerging studies to identify generative AI applications in software product management.Applications are aligned with ISPMA categories to show where they fit in the product-management framework.Source: §5 · p. 8, §5.3.1 · p. 9

5.1 Data Source and Search Strategy

• The search queried IEEE Xplore, ACM Digital Library, Google Scholar, EBSCOhost, and ProQuest Central using combinations of generative-AI and product-management terms.Search terms also covered idea generation, product design, customer insights, planning, code generation, UX, ethics, and customer-feedback analysis.Source: §5.1 · p. 9

5.2 Study Selection

• The review targeted English-language, full-text sources from 2016–2023, with database-specific inclusion and exclusion criteria.The criteria differ by database; Table 2 is identified as the summary of those criteria.Source: §5.2 · p. 9, Table 2 · p. 9

5.3.1 Applications of GenAI in Software Product Management

• The review maps identified applications to ISPMA categories, using the framework to organize where generative AI may be used in software product management.The framework includes activities such as positioning and product definition, requirements engineering, development execution, and customer support.Source: §5.3.1 · p. 9, Applications of Generative AI with… · p. 10

Applications of Generative AI with ISPMA Framework

• For market analysis, generative AI can analyze existing research and customer feedback, but cannot itself conduct primary research such as customer interviews.Karim et al. (2022) found it generated coherent, interconnected ideas from a given knowledge base; the cited study was in medicine.Source: Applications of Generative AI with… · p. 10
• With market and product analysis as input, generative AI can assist product definition, naming, and feature suggestions; Nguyen et al. (2021) reported promising product-description results.Customer insight applications include analyzing feedback and tailoring future responses to customer needs.Source: Applications of Generative AI with… · pp. 10–11
• Generative AI can support product and detailed requirements documentation, including Agile stories, and serve as a conversational product-knowledge system trained on product-related information.Managers may also use competitor information to generate positioning ideas and product titles.Source: Applications of Generative AI with… · p. 11
• Generative AI can produce UX design variations for prototyping and personalization, and support development execution through automatic code generation.Peng et al. (2023) reported that the Copilot group completed tasks 55.8% faster.Source: Applications of Generative AI with… · p. 11

5.3.2 Change Management and Strategy Models in AI Implementation

• Lewin’s Change Model organizes implementation into unfreezing, transition, and refreezing, including recognizing opportunities and risks, deploying systems and training staff, then monitoring and adjusting use.The described risks include output quality and accuracy, ethical concerns, and workflow integration.Source: §5.3.2 · pp. 12–13
• The McKinsey 7-S Framework applies strategy, structure, systems, shared values, skills, style, and staff to organizational alignment for generative AI.The paper notes that staff may need to review outputs and roles may need to be redefined to manage and maintain AI systems.Source: §5.3.2 · p. 13

5.3.3 Ethical Implications of Generative AI

• The paper identifies bias, data privacy, accountability, and misuse of generated content as ethical concerns requiring attention.It argues for robust ethical guidelines and regulatory measures for responsible use.Source: §5.3.3 · p. 13
• The Responsible Innovation framework brings anticipation, reflexivity, inclusion, and responsiveness to the consideration of generative AI’s benefits and risks.The paper connects these dimensions to anticipating output, ethical, integration, and intellectual-property risks; human review; stakeholder input; and adaptation after deployment.Source: §5.3.3 · p. 14
• The paper highlights tensions between generative AI and GDPR principles, including black-box limits on transparency and the volume of training data versus data minimization.It also discusses accuracy and the difficulty of limiting storage when systems retain data for continued learning.Source: §5.3.3 · p. 15
• The paper presents GDPR principles as a foundation for addressing generative AI’s ethical and practical challenges while protecting individual rights.It also describes the Responsible Innovation framework as a way to support responsible use while considering benefits and potential harms.Source: §5.3.3 · pp. 14–15
• Hacker et al. (2023) argue that the EU AI Act is ill-equipped for rapid advances in large generative models and advocate attention to content moderation under the Digital Services Act.They favor technology-neutral regulation and clearer terminology for actors in the AI value chain.Source: §5.3.3 · p. 15

6. Results

• Generative AI applications in software product management span market analysis, product definition, customer insights and support, requirements engineering, and development execution.Table 3 is identified as the summary of applications.Source: §6 · p. 16, Table 3
• Studies describe generative AI generating coherent ideas, creating interfaces, estimating user-story points, elaborating natural-language code, analyzing customer feedback, and supporting requirements and documentation work.The review says these uses can reduce manual effort in managing product requirements.Source: §6 · p. 16
• The review concludes that applications from product discovery to delivery can potentially streamline workflows, improve product quality and outcomes, and use resources more efficiently.The authors cite Karim et al. (2022), Houde et al. (2022), Siggelkow & Terwiesch (2023), Kim et al. (2021), Fu & Tantithamthavorn (2022), and Dam et al. (2019).Source: §6 · p. 16

Applications of Generative AI in Software Product Management

• Generative AI models can act as hypothetical customers in market research and may inspire research paradigms not restricted to human-subject research.The review also reports that larger models generate more coherent and interconnected ideas in specialized disciplines.Source: Applications of Generative AI in… · p. 16
• For product positioning and definition, generative AI can suggest names and features and generate product descriptions and titles.The passage reports demonstrated efficacy for e-commerce descriptions and title generation across industries.Source: Applications of Generative AI in… · p. 16
• Generative AI can support customer experiences by interpreting and integrating data, translating needs into requests, tailoring responses, and analyzing large volumes of feedback.The passage describes a positive feedback loop that can deepen customer understanding for more personalized service.Source: Applications of Generative AI in… · pp. 16–17
• Generative AI can automate requirements elicitation and analysis, produce natural-language requirements from structured data, and detect conflicts in SRS documents.These uses address both generating requirements and checking them for inconsistency.Source: Applications of Generative AI in… · p. 17
• In Agile development, generative AI can support sprint planning, backlog management, and estimation, including generating user stories from input data.The passage says this can save time, reduce human error, and improve user-story quality.Source: Applications of Generative AI in… · p. 17
• Generative AI can assist code and documentation generation, source-code comprehension, and creation of user-centered interfaces for rapid prototyping.The passage also describes generating UI/UX designs from user preferences and requirements.Source: Applications of Generative AI in… · p. 17
• Generative AI can enhance business decision-making, with younger users and males described as typically more open to adopting the technology.Source: Applications of Generative AI in… · p. 17

• The section asks what ethical challenges generative AI presents.The supplied passage does not include the remainder of the research question.Source: Applications of Generative AI in… · p. 18

strategies can be employed to mitigate these issues?

• Bias in training data can produce skewed or prejudiced outputs, making fairness and diverse training data important to avoiding potential harm or discrimination.Source: Ethical Implications · p. 18
• Extensive training-data requirements increase the likelihood of privacy infringement, while GDPR principles guide the collection, storage, and use of personal data.Source: Ethical Implications · p. 19
• Organizations are responsible for rectifying harmful or inaccurate AI outputs and preventing their recurrence (Cao et al., 2023).Source: Ethical Implications · p. 19
• GenAI can produce fabricated content and hallucinations, while intricate models’ black-box behavior limits interpretability and may undermine trust; verification and human review are needed.Source: Ethical Implications · p. 19
• Unresolved intellectual-property questions and confidential prompt risks call for protective contracts and confidentiality provisions.Source: Ethical Implications · p. 19
• Responsible Innovation principles and GDPR can guide mitigation, but regulatory challenges persist and the current EU AI Act fails to keep pace with rapid advances in large generative models.Source: Ethical Implications · p. 19

7. Conclusions

• GenAI can support product-management work from idea generation and market research to UI/UX design, requirements, development, and customer support.Source: §7 · p. 20
• Product managers must review AI-generated content for authenticity and accuracy, while addressing bias, transparency, privacy, accountability, ethics, and integration difficulties.Source: §7 · p. 20
• The conclusion calls for comprehensive guidelines, regulatory measures, and a reflexive, anticipatory approach to GenAI development and implementation.Source: §7 · p. 21

8. Future Research Recommendations

• Future research should develop models tailored to product strategy, planning, and feature prioritization, and explore integration of AI-generated stories, planning, and resource allocation with Agile methods.Source: §8 · p. 21
• Further directions include personalized AI assistance and evaluating long-term effects on workflow efficiency, cost-effectiveness, and product quality.Source: §8 · p. 21