Watch A2V think.
A short walkthrough of one synthetic organisation, from a confident board slide to a priced constraint and a leadership decision. It teaches how the method reasons — it is not the full diagnostic, and every organisation, person and figure below is invented.
The walkthrough
Halveston Group — distribution & field services
A fictional UK mid-market business, around 1,800 people, moving goods and running on-site engineering crews. It has bought AI tools, launched pilots, and published a policy. On the surface, things look busy and positive.
A guided walkthrough in ten short steps.
1 · The board slide
The quarterly board pack leads with AI: 11 pilots live · 3 enterprise tools rolled out · an AI policy published · "strong momentum". The slide is confident. Nobody in the room disputes it.
A2V starts by asking what that confidence rests on.
2 · Three questions
A2V puts three questions to the board — the ones AI activity is supposed to answer.
That's the honest answer for most organisations at this stage — and Halveston is no exception. There is activity everywhere, but no evidenced account of what value it has produced. Hold that thought.
A policy exists — but a policy is not a control. As we'll see, few people can recall what it says. "What could go wrong" is rarely answerable from a published document alone.
"What next" is the hardest of the three to answer with evidence, because it depends on the other two. Without knowing what's working and what's exposed, the next step is a guess dressed as a plan.
3 · The evidence
A2V triangulates three sources that can't all be wrong the same way: what people say, what systems record, and what we see.
"Adoption is broad — most of operations use the assistant daily." — Divisional director, interview
Licence telemetry: 22% of issued seats active in the last 30 days; usage concentrated in two teams.
On a depot shadowing visit, crews keep a private spreadsheet to redo the assistant's handover notes before trusting them.
Where the three diverge, the divergence is the finding. Here, "broad daily adoption" (said) does not survive contact with the telemetry (recorded) or the floor (seen).
4 · The perception gap
Leadership believes adoption is broad and working. The telemetry and the depot floor say it is narrow and quietly distrusted. That gap between the top and the floor is not noise to be averaged away.
The rule of the floor: where leadership perception and frontline evidence disagree, A2V scores toward the floor — and reports the gap itself as a finding. It is usually the most valuable insight in the engagement.
5 · Detecting theatre
Two items from the board slide look like progress but move nothing:
- A "concierge" chatbot pilot, live nine months, still described as "in pilot", with no owner and no metric. Pilot theatre
- A slide claiming "40% faster reporting" with no baseline taken before the tool went in, and no independent check after. Benefits theatre
Naming theatre kindly, with evidence, is often the moment the room goes quiet — and gets honest.
6 · Where the flow breaks
A2V walks nine flows that carry AI ambition to value. At Halveston, two are visibly blocked — at the boundaries no single function owns.
F1, the use-case flow: pilots enter but never exit — no scale decision, no kill decision. F7, the benefits flow: claims are made without baselines, so nothing is bankable. The depot handover delay sits right where these two meet.
7 · The So-What Chain
One finding, walked through the five steps that earn it a place in front of the executive team:
Finding
Depot crews redo AI handover notes by hand because the source notes live across three systems and none is trusted.
Flow blocked
F7 (benefits) — the time "saved" is spent re-checking, so it never converts to value.
Value capped
Illustrative arithmetic, not a real client figure: ~38 min per shift × crews in scope ≈ £180k–£320k a year of skilled time, confidence B.
Constraint
No single curated source of handover information — the binding constraint behind the leak.
Decision required
Fund a curated handover source and set a baseline — or accept the leak, explicitly.
8 · Evaluating a use case
Use cases in A2V are generated from evidence, never brainstormed, and each must survive a disqualification test: a named owner, a measurable baseline, and a real mechanism — not a wish to be seen doing AI.
That's the disciplined call. It has an owner, a baseline you can move, and one honest prerequisite — a curated handover source — which becomes a funded roadmap row, not a hidden assumption. It goes into the Now column.
A2V would disqualify this one — and say so openly. It is showcase-shaped: no owner, no baseline, and it exists to be demonstrated rather than to move a number. The disqualified list earns more trust than the approved one. You can change your choice, or carry it forward to see how the method records it.
9 · The leadership decision
The evidence is in and priced. A2V does not end with a recommendation — it ends with a decision the executive team owns.
Committed. The next step writes it into the Commitment Log — baseline, owner, date, evidence-of-done.
A legitimate choice — and in A2V it is minuted explicitly, not deferred silently. A recorded decline is a decision, and a success of the method.
10 · The Commitment Log
The executive team's decision is recorded.
An A2V engagement ends here — with a decision taken or explicitly declined, never with a report left on a shelf.
Prioritised Depot handover summariser
The use case you took forward, and why the method treats it that way.
| Outcome committed | Baseline today | Owner | Evidence-of-done |
|---|---|---|---|
| Cut depot handover time | 38 min average handover per shift (sampled) | Regional Operations Director | Handover minutes per shift, verified by Finance |
That is how A2V thinks. The real diagnostic walks your flows, your evidence and your money — and ends with your executives committed, or explicitly declining.