Cost calculator
What does our AI transformation actually cost, and when does it break even?
The full cost of an AI programme given your headcount, current maturity and ambition — with the return, the break-even point, and the lines quick estimates leave out.
- ≈ 3 min
- Free calculation · AI commentary on request
- You leave with: A line-by-line budget, the return and the break-even month
Your situation
The figures recalculate as you move. Nothing is sent anywhere until you ask for it.
The people who will have access to AI, not total headcount.
Salary, employer contributions and overhead — roughly 1.8× gross.
A shorter calendar does not remove work; it concentrates it.
Your raw assumption. The model then removes partial adoption and the unrecoverable share.
The result
The figures above are computed and free. If you also want them read back to you — what they mean, what is specific to your sector, where to start — that is produced on request.
When it breaks even
Benefit follows an adoption curve, not a straight line: slow at first while early users find their footing, then faster. The thick line is the cumulative net position.
The gap, axis by axis
Each axis costs differently. Governance is fixed senior time; skills is a per-head cost; integration carries the widest range because it depends entirely on the state of your systems.
- Gouvernance1.5 → 3.01.5 step to close
- Compétences1.5 → 3.01.5 step to close
- Intégration1.5 → 3.01.5 step to close
- Sécurité1.5 → 3.01.5 step to close
- Déploiement1.5 → 3.01.5 step to close
The breakdown
The lines marked “often omitted” are the ones a quick estimate leaves out. They are a large share of this total, which is precisely why AI budgets overrun.
| Line | Estimate |
|---|---|
| Set-up | |
Advisory and governance 41 j of senior advisory: usage policy, scoping, steering. | €60,750 €18,000 — €198,750 |
Technical integration 56 j of engineering: connecting data, securing it, getting to production. | €49,950 €13,500 — €176,400 |
| People | |
Training delivery 19 j of delivery to cover 80 people. | €26,040 €8,100 — €76,560 |
Participants' own timeoften omitted 1,800 hours of internal time at €75/h. Time not spent on the business — almost always missing from estimates, and frequently the largest single line. | €135,000 €72,000 — €252,000 |
Initial productivity dipoften omitted Hours lost learning the tool and producing worse work with it before producing better. Small per head, material at scale. | €36,000 €12,000 — €84,000 |
| Recurring | |
Licences (12 months) 8 advanced seats, 36 standard, 36 light. Billed on seats bought, not seats used. | €23,904 €11,520 — €39,840 |
Programme managementoften omitted Coordination, reporting, steering. Real, and invariably absent from the first estimate. | €16,409 €3,168 — €90,342 |
Change managementoften omitted Communication, a champions network, floor-level support. The line that gets cut first and causes the adoption failure that costs ten times more. | €13,674 €1,980 — €81,308 |
| Risk | |
Identified risks (weighted)often omitted 5 pitfalls detected for this profile, priced as probability × impact rather than flagged in the margin. | €137,593 €20,201 — €303,484 |
| Return | |
Benefit over 36 months 6.0 h/person/month in theory, cut to 60% active users, then to 50% genuinely recoverable value, and spread over a 6-month ramp. | €356,010 €124,603 — €756,520 |
Sensitivity
The net benefit depends on four assumptions at once. The one at the top is the one that, if wrong, changes the answer the most.
Each assumption is moved ±20% on its own, holding the others fixed. The longest bar is the one worth checking first — not necessarily the one that felt most uncertain.
What can go wrong
Detected from your profile, not from a generic list. Each pitfall states what triggered it, and its weighted cost is already included in the investment above.
Contingency provision
Low50 %What triggered it: The gap between the central and high estimate is €177k.
A budget set at the central estimate has, by construction, an even chance of being exceeded. That is not a flaw in the estimate; it is what a central estimate means.
What to doBudget the central estimate and hold the gap as a separate provision. An explicit provision that goes unspent is handed back; an overrun has to be explained.
- Impact if it lands
- €177,381
- Weighted cost
- €88,691
Integration targeted without engineering capacity
High50 %What triggered it: Target of 3/4 against current integration at 1.5/4.
Moving from a generic assistant to AI connected to real business data is an engineering project, not a licence purchase. With no internal capacity, the estimate slips on the least predictable part of the programme.
What to doScope one integration end to end before promising five. The first connector reveals the true state of the data; the ones after cost a fraction of it.
- Impact if it lands
- €60,000
- Weighted cost
- €30,000
It all rests on one person
Medium35 %What triggered it: Use is established (deployment 1.5/4) while collective skill is low (1.5/4) across 80 people.
That gap is the signature of capability concentrated in a few individuals rather than spread. Their departure takes the practice with it, and the training budget then re-buys what already existed.
What to doDocument what works — prompts, procedures, worked examples — and have those people run the training rather than an external provider. Transfer becomes a deliverable instead of a hope.
- Impact if it lands
- €25,000
- Weighted cost
- €8,750
Security debt on data
High9 %What triggered it: Security at 1.5/4 across 80 people with access to AI tools.
With no signed DPAs and no data partitioning, confidential or personal information passes through third-party services with no documented legal basis. The incident is not certain; its cost, if it happens, is known.
What to doSign DPAs with the chosen vendors, turn on the no-training settings, and update the processing register. A few days of legal and technical work.
- Impact if it lands
- €60,000
- Weighted cost
- €5,400
- Cost to prevent
- €9,000
No measurement of actual use
Medium60 %What triggered it: Deployment at 1.5/4: no structured usage tracking.
Without measurement there is no way to tell an active seat from a dormant one. The renewal is negotiated blind and defended poorly, because there is nothing to show but an invoice.
What to doTrack three numbers monthly from day one: active seats, self-reported hours, one documented use case per team. It costs nothing and it is what makes the budget defensible.
- Impact if it lands
- €7,920
- Weighted cost
- €4,752
The assumptions
Every figure above rests on these. The ones marked “assumption” have no external source: they are modelling choices, labelled so they can be argued with rather than mistaken for measurements.
- Headcount used 80Observed
- Entered directly.
- Active users 60 %Published
- Active use as a fraction of seats bought. Enterprise assistant deployments routinely report 30–60% sustained active use; the top of this range requires deliberate enablement, which is what the change-management line above pays for.
- Realisation factor 50 %Assumption
- The gap between hours saved and value realised. Twenty minutes saved across a fragmented day is often not recoverable as anything. This is the single biggest reason AI business cases overshoot, and it is why the tool never quotes raw hours × rate as a benefit.
- Ramp to steady state 6 monthsObserved
- Time from rollout to steady-state usage. Shorter where a single workflow is targeted, longer where use is left to individual discretion.
- Effort per maturity step 7.5 steps in totalObserved
- Effort to raise one axis by one maturity level, from i-avantage engagement data. Governance and Security are dominated by fixed senior time — writing a policy costs the same for 50 people as for 500. Skills is dominated by per-head training and therefore scales almost linearly. Integration is the volatile one: it depends entirely on the state of the systems being connected, which is why its range is the widest.
- Loaded hourly cost 75 €/hObserved
- Rate supplied by the respondent.
What the calculation does, what the AI does
Every figure on this page comes from a deterministic algorithm running in your browser: the same input always gives the same result, and nothing is sent to produce it. AI runs only afterwards, to write the commentary from those figures. It produces none of them — the format it fills has no numeric fields — and any sentence quoting a value the calculation did not produce is removed before display.
Where to go next
See all 19 tools →- Use-case prioritisation
Of all our ideas, which one first?
≈ 5 min · Your use-case list ranked, on a value/effort plane - Build or buy
Should we buy this tool or build it ourselves?
≈ 3 min · Both options costed and the point where building pays - Right-size licences
How many seats are we paying for nothing?
≈ 2 min · The saving available at renewal, and the tiered arrangement - Inference cost
What will the model calls cost once in production?
≈ 3 min · The monthly bill and its two-year trajectory - Internal team or external
From when is hiring cheaper than buying days?
≈ 3 min · The crossover month and how many hires are justified - Automate a process
Is this process worth automating, and how many months until it pays back?
≈ 3 min · A signable automation case, with a proposed decision