Whose Story Is It? — Working Agile with AI

The Question You're Actually Asking

Let me cut to it: you want to know if the work is still yours when AI helped create it.

The answer is yes. But only if you claim it.

Here's the uncomfortable truth most teams are avoiding: AI doesn't take ownership. AI doesn't have accountability. AI doesn't get fired when the code breaks or the story was wrong or the customer gets hurt. You do.

That clarity should actually make this easier. The work is yours because the consequences are yours. The question isn't philosophical — it's operational. Who's on the hook when this artifact hits production?

Why This Feels Harder Than It Should

I watch teams tie themselves in knots over AI ownership, and the real issue usually isn't legal or technical. It's identity.

Think about it from the Frame of Reference of the person doing the work. If you've spent ten years becoming an excellent technical writer, and now AI drafts documentation in thirty seconds, what does that mean about your contribution? If you're the analyst who built your reputation on thorough research, and AI summarizes sources faster than you can read them, where does your value come from now?

This isn't resistance. It's a legitimate question about professional worth. And if you don't surface it, it surfaces itself — in delayed reviews, "quality concerns" that are really control concerns, and passive non-adoption that nobody names directly.

The person who reviews AI output and says "this is ready" is taking professional responsibility. The person who catches the hallucination before it becomes a customer problem is doing work that matters. The person who improves the prompt so next time the draft is better is building organizational capability.

That's ownership. It's just ownership of a different set of activities than before.

The Simple Rule That Actually Works

Not "reviewed by." Not "assisted by AI." Your name, your accountability.

When a developer signs off on AI-generated code, they're saying: "I understand what this does, I've tested it, and I'm responsible if it breaks." When an analyst approves AI-drafted acceptance criteria, they're saying: "This reflects what the customer actually needs, and I'll own it if it doesn't."

That signature is the moment the work becomes yours. Everything before that is drafting. Everything after is ownership.

What this looks like in practice:

• AI drafts the user story. Human reviews, edits, adds edge cases, and signs. The human owns it.
• AI generates test scripts. Human validates coverage, runs them, confirms results, and signs. The human owns it.
• AI summarizes the sprint retrospective. Human reviews for accuracy, adds context, and signs. The human owns it.

The AI did work. The human took responsibility. That's the distinction that matters.

What You're Really Afraid Of (And Whether You Should Be)

Let's name the fears I hear most often:

"What if AI made an error I didn't catch?"

Then you made an error. That's always been true. If you approve a document with a typo, you own the typo. If you approve code with a bug, you own the bug. AI doesn't change this — it just means you're reviewing AI output instead of writing from scratch. The accountability is the same.

The real question is: are you reviewing carefully enough? AI makes certain kinds of errors (hallucinations, plausible-sounding nonsense, missing context) that are different from human errors. Learn the patterns. Check the facts. Verify the logic. That's your job now.

"What about intellectual property? Who owns AI-generated content?"

This is a real legal question, and it varies by jurisdiction and AI vendor terms of service. I'm not a lawyer, and you should talk to yours if this matters for your context.

But here's what I know from the implementation side: legal ownership and operational ownership are different things. Even if the legal questions take years to resolve, your team needs to know today who's accountable when something goes wrong. That's the human who signed off.

"What if my contribution doesn't feel meaningful anymore?"

This one's the hardest because it's about identity, not policy.

Here's what I've seen: some people who struggle with AI are the ones who defined their value by production volume. "I wrote 50 user stories this sprint." "I documented the entire API." When AI can produce volume faster, that metric stops meaning what it used to.

The people who adapt fastest are the ones who defined their value by quality and judgment. "I caught the requirement that would have caused a production incident." "I knew this approach wouldn't work for our context and redirected before we wasted two weeks." AI can't do that. You can.

What Leaders Need to Know About Ownership

Ownership doesn't start with the team. It starts with leadership.

IMA's AIM research, built over 40+ years of studying what makes implementations succeed, shows a clear pattern: what leaders say has 1x impact on behavior change. What they visibly do has 2x impact. What they reward and recognize has 3x impact.

This means the most powerful signal about AI ownership comes from what leaders do themselves — not from policy announcements or training programs.

When leaders use AI visibly, talk about their experience (including what didn't work), and recognize team members who are building good AI practices, adoption accelerates. When there's a gap between what's announced and what's modeled, people notice. They take their cues from what gets rewarded, not what gets announced.

What leadership ownership looks like in practice:

• Using AI to prepare for your own meetings and saying so: "I used Claude/Gemini/ChatGPT/etc. to draft this agenda — let me know if I missed anything."
• Sharing your learning curve openly: "This prompt didn't work the first time. Here's what I changed."
• Updating performance goals to reflect new expectations: "Q2 objectives now include AI tool adoption for documentation."
• Recognizing the right behaviors publicly: "Sarah caught a significant error in AI-generated documentation before it shipped. That's the quality standard we need."
• Having direct conversations with people who are struggling: "Help me understand what's getting in your way."

The challenge for most leaders is that AI adoption asks them to be visibly learning something new — which can feel uncomfortable. But that visible learning is exactly what gives others permission to do the same.

Making It Real: The Artifacts That Matter

Stop treating AI metadata like optional documentation. Start treating it like the audit trail it is.

Every AI-assisted artifact should answer these questions:

• Who owns this? (Name, not role) — Accountability when something goes wrong
• What AI was involved? (Model, date) — Reproducibility and understanding limitations
• What prompt produced it? — Learning and improvement over time
• What instructions were used? — Context that shaped how the AI responded
• What data sources were used? — Verifying accuracy and identifying bias
• What human review happened? — Validation that judgment was applied
• Who signed off? — The moment ownership transferred

The data source and instructions questions matter more than most teams realize. AI models can pull from training data, web searches, uploaded documents, or combinations of all three — and system instructions or custom settings shape how that information gets used. When something turns out to be wrong six months later, you need to know: did the AI hallucinate this, did it come from a source that was outdated or incorrect, or did the instructions steer it in the wrong direction? The answer changes how you fix the problem and prevent it from happening again.

The Collaboration Question

Here's a tradeoff most teams aren't tracking: AI can make you more efficient and more isolated at the same time.

When AI handles status updates, summarizes meetings, and drafts communications, you save time. But you also lose the informal knowledge transfer that used to happen in those activities. The "I overheard you mention X" moments. The context that doesn't fit in a summary. The relationship-building that happens when humans talk to each other.

I've seen teams celebrate AI efficiency gains and then wonder, six months later, why nobody knows what anyone else is actually working on. A deadline slips badly and everyone's surprised — but no one should have been surprised if they'd been talking.

The practical balance:

• Use AI for collection, not for connection. AI gathers the status updates; humans discuss what matters.
• Reserve live time for problems, decisions, and the things that require actual conversation. It's not about the formal stand-up — it's about the "meet after." The hallway conversation, the quick huddle, the "can I grab you for a sec?" moments where real context gets shared and problems get solved before they blow up. That's what AI efficiency can accidentally squeeze out if teams aren't paying attention.
• Track collaboration signals: How often do people reach out to each other? How much time goes to problem-solving vs. reporting? If those metrics fall, your efficiency gains are costing you something important.

Own AI Outputs — IMA Worldwide

AI is a tool. You are the owner.

The work is yours when you claim it — when you review it, judge it, improve it, and put your name on it. The work is yours because the consequences are yours. That's always been true. AI doesn't change the principle; it just changes what you're reviewing.

If that feels like a loss of something — the creative act, the production volume, the identity you built around a certain way of working — I hear you. That disruption is real. But the answer isn't to pretend AI isn't changing things. The answer is to get clear about what ownership means now and step into it.

Your leaders need to model this. Your team needs to practice it. Your artifacts need to reflect it. And you need to decide: are you going to own this work or not?

Quick Checklist

For You Personally:

• Can you explain what you contributed to every AI-assisted artifact with your name on it?
• Do you know the error patterns of the AI tools you use?
• Have you caught an AI mistake before it shipped? (If never, you're not reviewing carefully enough.)

For Your Team:

• Every AI-assisted artifact has a named human owner
• Ownership transfers at signature, not at draft
• Prompts and AI metadata are tracked for learning
• Review quality is measured, not just review completion

For Your Leaders:

• Leaders use AI visibly and talk about their experience
• Performance goals reflect AI adoption expectations
• AI adoption behaviors get recognized publicly
• Non-adoption gets addressed directly (not punitively, but clearly)