Cost an AI decision, don't imagine it

What does our AI transformation actually cost, and when does it break even?

Counts your teams' own time, the adoption ramp, and the share of saved hours that never converts into value.

Of all our ideas, which one first?

The data-readiness gap enters the effort — which is where the difference between a quick win and a two-year project hides.

Should we buy this tool or build it ourselves?

Includes the production-hardening multiplier — the factor of two no initial estimate contains.

How many seats are we paying for nothing?

The most concrete money in the suite: a figure to put on the table at the next renewal.

What will the model calls cost once in production?

Separates input from output, because answer length is what surprises finance in month three.

From when is hiring cheaper than buying days?

Counts the fully-paid ramp-up and the recruitment — the two reasons the crossover lands later than expected.

Is this process worth automating, and how many months until it pays back?

Counts human review at the full rate and the quality gap in both directions: where an 'obvious' automation stops being one.

What is going to go wrong in our programme, and what will it cost?

Says nothing when nothing is wrong. A healthy profile fires no rule, and that silence is the result.

What is ungoverned AI use already costing us?

Sets that cost against the cost of closing the gap. It is the only comparison that moves a budget.

Is our system high-risk, and what does that oblige us to do?

A rule engine because the Regulation is one. Orients before the legal opinion; does not replace it.

Which vendor should we trust with our data?

Without a DPA, price does not matter: the data cannot lawfully leave. A spreadsheet lets it through; this does not.

Our teams already use AI: what rules do we give them, in writing, this week?

Every cell of the data table is decided by your answers, with its condition — and a stricter stance can never open one. A downloaded template decides nothing.

In what order, and over how long?

The order is the dependency graph, not a preference: what is built without governance gets rebuilt.

How many sessions, over how long, for what budget?

Derives the duration from your delivery capacity, not your budget. It is frequently the number that changes the plan.

The hours saved — how many people is that, and what does it become?

Refuses to book FTEs as a saving without saying through which departures. That is what makes the case signable.

How much does our AI use emit, and where should it run?

Honest about per-token uncertainty, solid on grid intensity — and that is what dominates. The French grid is an argument.

Is our AI pilot big and long enough to prove a gain — or only to produce impressions?

Separates the spread between people from the spread task to task: measuring more tasks is often cheaper than recruiting more people.

Which holding carries the most AI value — and does it arrive before exit?

Ranks on the holding window, not on run-rate: a company 18 months from exit captures almost nothing.

What does the engagement for this client look like, in days, quote and margin?

Built on the public calculator's constants: a client who used the tool recognises the same programme in the proposal.