AI is your fastest drafter — but only if you brief it right.
More and more product teams are leaning on ChatGPT to help with Agile artefacts. The idea is simple: offload the drudgery of drafting user stories, subtasks, and acceptance criteria so you can focus on higher-leverage work. But for many professionals, especially those under time pressure, the promise quickly turns into frustration.
They expect acceleration. What they get is a rewrite.
Scenario: When Speed Kills Clarity
A delivery-side business analyst is tasked with drafting three related user stories. Each one touches a different layer of complexity — one involves backend logic, one maps a user flow, and one hinges on system integration.
They all belong to the same initiative, and she’s under pressure to get them ready for sprint planning. To save time, she turns to ChatGPT, but although well-intentioned, her prompt is vague:
“Can you write three user stories for this feature?”
The output is fast, but flawed.
- Roles are muddled
- Dependencies ignored
- Edge cases imagined
So she tries again. Reframes the request. Refines the prompt. Copy-pastes snippets from other stories. Adds bullet points. Adjusts formatting. Pastes in the epic. Trims the responses. Rewrites the acceptance criteria by hand.
Eventually, she gets what she needs, but it took longer than writing from scratch.
The Context Pack Comes First
The core mistake wasn’t the prompt — it was the absence of scaffolding.
Professionals often assume AI can intuit context from a few lines of text. But unlike your team, ChatGPT wasn’t at the last sprint review. It doesn’t know your constraints, your roadmap, or your internal shorthand. When you ask it to draft a story without giving it anything to stand on, it pulls from training data — not your product.
A context pack fixes that.
Think of it like a blend between a Product Initiation Document and a team onboarding guide. It should include working assumptions, user roles, naming conventions, known constraints, and example artefacts. Nothing flashy — just structured reality. When used with ChatGPT, this becomes a silent co-pilot, guiding the model away from hallucination and toward alignment.
Even without AI, a good context pack improves onboarding, stakeholder communication, and team resilience. With AI, it turns garbage prompts into gold.
Templates Aren’t Just for Teams — They’re for Training the Model
The best way to shape AI output isn’t to explain what you want — it’s to show it.
If your team already uses a template for writing user stories or epics, that template should be fed to the model before you ask it to generate anything. If you don’t have one, build one. AI learns from pattern recognition. It doesn’t care about “best practice.” It mirrors what it sees.
If your stories follow the INVEST principles or include BDD-style acceptance criteria, that’s exactly what ChatGPT needs to see. If your team uses bullets instead of paragraphs, or splits acceptance criteria by role, include those examples. Otherwise, you’re asking the model to invent structure — and that’s a guaranteed way to waste time.
Formatting Is How You Stop the AI From Guessing
Even when the content is technically correct, poor formatting can make the AI output unusable.
If you want UK English, say so. If you want no fluff, short paragraphs, and strict use of defined terms — define them. ChatGPT doesn’t just respond to what you say; it responds to how you say it. Structure is its instruction set.
Without formatting guidance, your prompt is an open invitation to guess. And no professional wants to spend time cleaning up a guessed document. Think of formatting like linting for language — it prevents rot before it sets in.
Define it once. Reuse it often.
Conclusion: Tactical Is Practical
This isn’t about “prompt engineering.” It’s about operational clarity.
ChatGPT doesn’t fail because it’s dumb. It fails because it doesn’t know what matters to you — unless you show it. That means putting in place a context pack that captures your product’s reality. It means sharing templates so the model knows your structure. It means formatting your brief so the AI can mirror your clarity, not your confusion.
The tactical way to write stories with AI isn’t to hope it understands. It’s to train it like a teammate.
Give it your tools. Then watch it start pulling its weight.
Tactical Takeaways
How to Make AI Actually Useful
-
Build a context pack before you brief AI
Capture your product’s reality once — then reuse it across tools, teammates, and models -
Use artefact templates as training data
Show ChatGPT how your team writes, formats, and scopes — don’t make it guess -
Define formatting rules in your prompt
Spelling, tone, bullet styles — the more you specify, the less you edit later -
Refine structure, not just wording
Most AI output problems are structural, not semantic — fix the skeleton, not the skin -
Optimise for reusability, not novelty
If it works once, it should work again — treat every prompt as an asset, not a one-off -
Stop relying on the model to ‘get it’
AI isn’t a teammate — it’s a mirror. Clarity in, clarity out
