Illustrative example, Fictional data, Not a client result
A first board map for a 300-person distributor
What a first map can look like after early Consult hours: where AI already runs, what is known and how we know it, what is still a guess, and one first move worth making. The company and every figure are invented.
- Before
- Each pilot owner reports a success story. There is no list of what runs where, no baseline, and no agreed rule for what needs approval.
- After
- One page the COO can walk the board through: each area’s AI use, the evidence behind each claim with its date, the questions still open, and one reversible first move with an owner and a measure.
Inspect the sample
01 Input
The COO of Calder Wynn Supply, a fictional 300-person industrial distributor, has six weeks until the board asks for an AI strategy. Finance runs an invoice-matching pilot, sales bought two writing tools, support is testing a chatbot, and nobody has compared them.
How it moves
- Gather what already exists List every AI tool, pilot and vendor promise by department, with its owner and the data it touches. Nothing is scored yet.
- Date every claim Each statement about results gets a source and a date, or it is marked as an open question. A vendor’s case study stays a claim until your own numbers support it.
- Rank the open questions Order the questions by how much the answer could change the plan, against the cost and time of finding out.
- Choose a first move you can undo Pick one move with a baseline, an owner, a measure and a named approver. Moves that cannot be undone wait for the board.
02 Output
| Area | AI in use today | What we know (source, date) | Open question | Next move and approver |
|---|---|---|---|---|
| Finance | Invoice-matching pilot used by two clerks | Pilot owner reports fewer weekly exceptions (pilot log, 12 Sep 2026; not yet checked) | Does accuracy hold at month-end volume? | Run one month-end in parallel; Controller approves |
| Sales | Two writing tools, bought separately | No usage data yet (no source) | Which tool do reps actually use, and for what? | Pull 30 days of usage; Sales VP decides |
| Customer support | Chatbot test on the help page | Vendor deck claims high deflection (Aug 2026; unverified) | How many past tickets would it answer correctly? | Score 200 past tickets offline; Support lead approves |
| People and HR | Staff using personal chat accounts | Staff survey mentions it (Sep 2026) | What company data has been pasted in? | Interim use rule; COO and counsel approve |
| IT and security | Connected AI apps with no inventory | Admin console lists 23 third-party apps (IT export, 3 Oct 2026) | Which apps can read customer data? | Access review; IT lead signs off |
03 Review
Where a person decides
The COO reviews every line before the map leaves the room. A claim without a source stays an open question, and every next move names the person who must approve it before anything changes.
Limits of this example
Fictional company and figures. A real map depends on what your people can show and how quickly answers come in; a single block may cover one or two departments, not the whole company. Not legal or compliance advice.
See whether this fits your work, and what it would take.