The In-House AI Person — Scoping Card & Rollout Map · Resource

Keep this open while you scope your next AI project at work. Use it in two passes. First the scoping card: it forces the order that makes the project pay off — name the constraint, pick one number, only then build. Then the rollout map: four steps you tick off as you grow into the role from where you already sit.

The one rule everything runs on: constraint first, KPI second, build third. If you skip straight to building, you can spend a week saving twenty minutes on a task nobody was waiting on.

Part 1 — Project-scoping card

Fill this in before you open a single tool. If you can't finish the card, you're not ready to build yet.

1. The constraint — what actually slows the business?

Write one sentence. Not the annoying task; the bottleneck.

The constraint I'm targeting: ________________________________
If I fix it, the business gets: (circle one) faster / more money / stops leaking money somewhere
Who feels this pain today, and where downstream does it show up? ________________

2. The KPI — which number, by how much?

One number. This is the project's North Star and the sentence nobody else in this field bothers to say.

The single number I'm trying to move: ________________________________
Where it sits today: __________ → where I want it: __________
How I'll measure it (the source of truth for this number): ________________

3. The build — only now do you scope it

The smallest version that could move that number: ________________________________
One corner of the company I can test it in (one team, one process): ________________

Sanity questions — catch the trap before you build

Run these four before you commit. A "no" or a shrug means go back to the card.

[ ] If this works perfectly, does a constraint get smaller — or did I just automate a task?
[ ] Can I name the one number it moves, and say where that number lives today?
[ ] Would someone in the business notice if this fix existed — or was nobody waiting on it?
[ ] Can I point to the result afterward and say what it moved, for whom?

Part 2 — Four-step rollout map

You can start step one tomorrow without telling anyone. Work down the list; each step earns the next.

Step 1 — Audit your role through the constraint lens

[ ] Listed the parts of my job that gate revenue, slow the team, or cause real pain downstream — not just the repetitive ones.
[ ] Crossed off the "easy win" tasks that wouldn't actually change anything.
[ ] Wrote the specific number I'd move next to each item that's left.
[ ] I now have a short project list of consequential fixes, not easy ones.

Step 2 — Ship one small project with proof

[ ] Picked one item and scoped it with the card above (constraint → KPI → build).
[ ] Ran it in one corner — small enough to actually test and watch.
[ ] Captured the result as my first case study: what it moved, where, by how much.
[ ] Have somewhere to point to it (a note for my manager, the team, a simple portfolio).

Step 3 — Spot the recurring patterns

[ ] After a few small builds, named the handful of problems that keep showing up across the business.
[ ] Practiced naming the real problem in a situation before reaching for a tool.
[ ] Noticed the flip: coworkers start coming to me instead of me pitching them.
[ ] Checked my own judgment — caught at least one case where the model's confident answer was wrong.

Step 4 — Formalize the role with evidence

[ ] Gathered enough proof that the company clearly needs this done full-time.
[ ] Took the evidence (not a job posting) to my manager or leadership and proposed the role myself.
[ ] If they're not ready yet, made sure I'm the first name that comes to mind when they are.

Your quiet next step is the top of the card. Sit down this week and find the single constraint in your role that, if it broke loose, would actually make the business faster. Write the number you'd move. That line is where the role begins.