Back to the library

Key question New

AI-processed · Author-approved

How do managers take ownership of AI-assisted work, and when does disclosing AI use help or hurt credibility?

When Is It Still My Work? Generative AI, Defensibility, and Disclosure in Software Product Managers’ Work
Phong Trang, Tran Dieu Hieu Nguyen, Soumitra Chowdhury · International Conference on Information Systems (ICIS) · 2026
Illustration for: How do managers take ownership of AI-assisted work, and when does disclosing AI use help or hurt credibility?

Loads Spotify's player; Spotify may set cookies. This device remembers your choice.

Key findings

  • Managers commonly rely on an 'Ask-Tweak-Share' routine, using the critical 'Tweak' phase to embed proprietary context, technical constraints, and business judgement into AI-generated drafts.
  • Artefact-level defensibility is established during modification before sharing, ensuring the author can answer detailed stakeholder questioning regardless of the tool used.
  • Defensibility and disclosure-mediated visibility can decouple, as some managers concealed AI assistance in substantially reworked documents to prevent peers from discounting their effort.
  • The study proposes that disclosure is more likely to encourage team experimentation when colleagues read the work as substantially reworked, and to draw a penalty when they read it as a shortcut.
  • In environments lacking formal governance, local team norms and peer reactions informally determine which AI practices become acceptable and openly discussable.

What it means for you

CIO / IT Executive

Where AI use runs ahead of policy, build guidelines on existing team norms. Encourage disclosure of substantially reworked work, and protect staff, especially early-career staff, from being judged as low effort when they disclose AI use.

IT Manager

Review team routines to ensure that every AI-assisted draft undergoes the critical 'Tweak' phase—embedding proprietary context, technical constraints, and business judgement—before it is shared with stakeholders.

Business Strategist

Before sharing an AI-assisted roadmap or rationale, check it against the decisions, constraints and history colleagues are likely to ask about.

Researcher

Test whether defensibility and visibility decouple in a setting with formal AI governance, and examine whether seniority and local team norms change how disclosure is received.

Policymaker

When writing guidance on AI disclosure at work, consider that the same disclosure can be read as openness or as low effort, depending on how much the work was reworked.

Summary

This qualitative study investigates how software product managers integrate generative AI into their everyday coordination routines in the absence of formal corporate policies. Through semi-structured interviews and screen-sharing sessions with 21 product managers at an automotive software firm, the researchers examine how professionals produce defensible work and navigate disclosure decisions. The analysis outlines how managers evaluate AI drafts against organizational realities and social expectations before sharing deliverables. The findings come from one automotive software organisation and are offered for analytical transfer to similar settings.

Problem

Managerial work relies heavily on producing shared deliverables, such as product roadmaps and strategic rationales, which carry professional accountability and require substantiation. Generative AI makes authorial provenance opaque, leaving managers uncertain about how to justify AI-assisted outputs as their own work when scrutinized by peers. Furthermore, professionals face dilemmas regarding whether disclosing AI involvement demonstrates innovation or risks damaging their credibility.

Transcript
Host: Welcome to A.I.S. Insights — powered by Living Knowledge. I am Anna Ivy Summers. Today we are examining a study titled "When Is It Still My Work? Generative AI, Defensibility, and Disclosure in Software Product Managers’ Work" by Phong Trang, Tran Dieu Hieu Nguyen, and Soumitra Chowdhury, presented at the International Conference on Information Systems in 2026.

Expert: Glad to be here, Anna. In short, this qualitative study investigates how software product managers integrate generative AI into their everyday coordination routines when their company lacks formal corporate AI policies. It looks closely at how professionals produce defensible work and decide whether or not to disclose AI assistance.

Host: Let us look at the core problem. Product managers carry significant responsibility for deliverables like roadmaps and strategic rationales. What makes generative AI tricky in this context?

Expert: Managerial work relies heavily on these shared documents, which require personal accountability and detailed substantiation. Generative AI makes authorial provenance opaque. When an algorithm drafts text, managers face uncertainty over how to justify the output as their own work when questioned by colleagues. They also face a dilemma: does disclosing AI use demonstrate innovation, or does it risk damaging their professional credibility?

Host: How did the researchers investigate this tension?

Expert: The authors examined twenty-one product managers at an automotive software firm. They conducted semi-structured interviews combined with screen-sharing sessions to observe how managers actually prompted models and refined drafts. Because the data comes from a single automotive software organization, the researchers present their findings for analytical transfer to comparable technical settings.

Host: What were the primary findings regarding how managers handle these tools daily?

Expert: The researchers observed a three-step routine they termed Ask-Tweak-Share. While the "Ask" step generates the initial text, the "Tweak" step is where the critical work happens. Managers use this phase to inject proprietary technical constraints, organizational context, and strategic business judgment into the AI drafts.

Host: So managers do not simply pass along the raw AI output.

Expert: Exactly. This editing process establishes what the study calls artefact-level defensibility. By modifying the text, managers ensure they can answer probing stakeholder questions about any detail in the document, regardless of the underlying tool used to create the first draft.

Host: Once managers achieve that defensibility, do they openly disclose that AI helped them?

Expert: Not necessarily. The study found that defensibility and disclosure can decouple. Even after substantially reworking a document, several managers chose to conceal AI assistance because they feared peers would discount their effort or assume they took an unearned shortcut.

Host: What dictates whether disclosing AI use is seen positively or negatively?

Expert: The study suggests disclosure encourages team experimentation when colleagues perceive the deliverable as thoroughly reworked. Conversely, it draws reputational penalties when peers view it as an unvetted shortcut. In the absence of official corporate rules, local team norms and peer reactions informally govern which AI practices become acceptable.

Host: Moving to the business takeaways, what can practitioners and leaders do with this knowledge?

Expert: First, business leaders should realize that policy vacuums force employees to manage social and reputational risks on their own. Second, organizations should shift their evaluation criteria toward artefact defensibility—focusing on whether an output solves the problem and withstands scrutiny, rather than trying to police the drafting process. Finally, setting clear disclosure standards can remove the stigma of using AI, enabling teams to openly share practical prompting and refinement techniques.

Host: Focusing on contextual refinement makes the work defensible, while clear norms allow teams to collaborate without fear of reputational loss. Thank you, Alex, for sharing these insights, and thank you to our listeners for joining A.I.S. Insights — powered by Living Knowledge. Until next time.
Artificial Intelligence & Automation Information Systems & Management Human-Computer Interaction Data Analytics & Big Data Digital Transformation & Strategy Cybersecurity & Privacy Cloud & Infrastructure E-Commerce & Digital Platforms Operations & Supply Chain Sustainability & Green IS

Generative AI, Software Product Management, Managerial Work, Accountability, AI Disclosure, Work Routines, artefact-level defensibility, disclosure-mediated visibility