PM Mapped
Working with AI STAKEHOLDER · User Stories and INVEST

User Stories and INVEST with AI

AI is genuinely good at the mechanical parts of user stories — drafting them in INVEST shape, splitting epics into shippable slices, checking the form. That frees you to focus on the part it can't do: deciding whether a story is the right thing to build at all.

← Back to the User Stories and INVEST tool
1How to use AI for this

A step-by-step way to work

1

Describe the need, not the solution

Give AI the user need and context; let it propose the story rather than dictating the wording.

2

Ask for INVEST-shaped output

Request stories that are independent, negotiable, valuable, estimable, small, and testable — and ask it to note where a story violates one.

3

Use it to split

Hand it an epic and ask for several ways to slice it into independently shippable stories. AI is good at spotting seams.

4

Check, don't trust

Review each story for whether it reflects a real user need and a scope you actually want — AI optimises form, not priority.

5

Keep acceptance criteria separate

Draft stories here; do criteria as their own step (see the Acceptance Criteria guide).

2Worked examples

Real prompts, and what good output looks like

Drafting a story from a need
Your prompt

Turn this into an INVEST user story: 'New users often abandon setup because they don't know which integrations they need first.' Persona: a new admin.

What good output looks like

Story: As a new admin, I want recommended integrations highlighted during setup, so that I can get value quickly without researching every option.

INVEST check: Independent ✓, Valuable ✓, Testable ✓. Small? Borderline — 'recommended' implies a recommendation rule that may be its own story. Consider splitting the recommendation logic from the highlighting UI.

The useful bit: AI flags that 'recommended' hides a second story. That seam-spotting is where it adds real value.
Splitting an epic
Your prompt

Split this epic into independently shippable stories: 'Users can export their data.'

What good output looks like

1) As a user, I can export my data as CSV, so I have a portable copy. (Ship first — simplest, highest demand.)

2) As a user, I can choose which date range to export, so I'm not forced to take everything.

3) As a user, I can export as JSON, so I can move data programmatically.

4) As a user, I receive a download link by email for large exports, so big jobs don't time out.

Each ships value on its own; #1 is the thin slice that proves the path.

Notice: it ordered by thin-slice-first. Whether that order matches your priorities is your decision, not the AI's.
3Copy-paste template

A prompt you can reuse

Fill in the highlighted parts and paste it into your AI tool of choice. Edit the output — it's a starting point, not a finished answer.

Reusable prompt
Help me write user stories in INVEST form. Don't add acceptance criteria — I'll do those separately.

The need / context:
[describe the user need, the persona, and the problem — not the solution]

Please:
1. Draft the user story as: As a [persona], I want [capability], so that [benefit].
2. Run an INVEST check and flag any letter it fails, especially if it's too big to be 'Small'.
3. If it's too big, propose 2–4 ways to split it into independently shippable stories, ordered thin-slice-first.
4Common pitfalls

What AI gets wrong here

Solution-shaped stories

AI will happily write 'As a user I want a dropdown…' — baking in a solution instead of a need.

Do this instead: Keep the story about the capability and benefit; let design decide the dropdown. Reject solutioned stories.

Plausible but mis-prioritised splits

AI orders splits by what's simplest, not by what matters most to your users or strategy.

Do this instead: Treat the ordering as a suggestion and re-sequence against your actual priorities.

Fake 'Small'

AI may call a story small when it hides a rule engine or integration.

Do this instead: Probe any story with words like 'recommended', 'smart', or 'automatically' — they often conceal a second story.
The judgment that stays yours

AI shapes the wording and proposes the splits — the craft of the story — but whether a story is worth building, and in what order, is prioritisation, which depends on strategy and evidence it doesn't have. Let it do the mechanics; you own the decision about what goes into the sprint.