Versatil
A free instrument for the person accountable

How much of your AI runs where you cannot answer for it.

The average enterprise spends 2,068 $ per employee per year on AI. The median spends under 200 $. An average with ten times its median describes nobody, so this positions you instead: your industry, your headcount, and the one split no published source carries.

Every published figure is sourced at the bottom of this page. Nothing here is a Versatil measurement.
Step 1
Your industry sets the intensity.

AI spend per employee varies by a factor of five between sectors, because knowledge work has more automatable surface per head than capital-intensive work. Pick yours.

Step 2
Your operation sets the exposure.

The third input is the one that matters and the one nobody publishes: how much of your AI work produces something that leaves the building. An internal error costs rework. An external error becomes a contract, a filing or a claim, and redoing the work does not undo it.

Headcount in scope.
internal 40%external 60%
Client deliverables, sales commitments, regulated filings, advice given. This is what the audit measures precisely.
ungoverned65%
Default is the published 2026 shadow AI adoption rate. Only 37% of organisations have a policy to detect it.
AI budget
2.07 M$
headcount times your sector's intensity
Running ungoverned
1.34 M$
outside any policy you can point to
Ungoverned and external
807 k$
where a wrong output cannot be recalled
Leaking, per year
1.2 M$
waste plus likely overrun, no budget line carries it
Where your sector sits
AI spend per employee per year, published 2026 figures
Your ungoverned surface, by where the output lands
the split you set in step 2, applied to the ungoverned share
internal
external
Internal, failure mode is rework External, failure mode is liability
What leaks inside the budget you already approved
applying the published waste and overrun rates to your figure
Shadow duplication, wrong-sized models, unused licences, at 30 to 50%
Likely overrun, since 63% of companies exceed by 30% or more in year one
Leaking every year, with no line carrying the name
Step 3
Your departments set where it lands.

Adoption in production is published for four departments and not for the other three. The three without a figure carry no bar, because a cell we have not measured is worth less than an empty one. Each row is per 100 people in that department, at the intensity and governance gap you set above, so nothing here assumes how your headcount is distributed.

External, the output leaves the building Internal, the failure stays in house RRegulated, a wrong output can become a claim

Read the order. The three most adopted departments are customer service, IT operations and marketing, and two of the three are external: adoption ran ahead precisely where a wrong output cannot be recalled. HR is the uncomfortable one, internal and heavily adopted for screening and regulated, which is the combination that produces a claim rather than a rework.

What this cannot tell you
The consequence multiplier, which is the number that decides everything.

This instrument apportions your spend. It deliberately does not multiply external exposure by a risk factor, because no honest source publishes one and inventing it would make every other number on this page suspect. What a wrong external output actually costs in your operation depends on your cycle, your gates and who signs. Four weeks measures it. See the audit.

Every published figure this page uses
2 068 $average AI spend per employee per year, 2026. Median under 200 $, top decile 2 800 $ and above.Rize, AI spend per employee benchmark
3 470 $professional and business services, the highest sector intensityAIStackHub, AI spending by industry
672 $manufacturing, the lowest, roughly five times below professional servicesAIStackHub, AI spending by industry
2 200 to 3 200 $finance and insurance. Two published sources disagree, so the range is carried rather than resolved.AIStackHub and Rize, in disagreement
65%shadow AI adoption in 2026. 78% of AI users at work bring unauthorised tools, 27% have entered confidential data into public AI.Airia, shadow AI statistics 2026
37%of organisations have any policy to manage or detect shadow AI useAiria, shadow AI statistics 2026
30 to 50%of total AI spend lost to shadow duplication, wrong-sized models and unused licences. Unused licence rate at a record 51%.Zylo and Cledara, SaaS and AI spend benchmarks
63%of enterprises exceed their AI budget by 30% or more in year onepublished 2026 enterprise AI spend research
4.63 M$average cost of a breach involving shadow AI, of which 670 k$ is attributable to the shadow AI itselfIBM Cost of a Data Breach, via Airia
40%+of agentic AI projects cancelled by end 2027, for escalating costs, unclear business value and inadequate risk controls. Mechanism named: agent sprawl.Gartner, June 2025 press release
17% to 64%organisations with agents deployed today, versus those planning production by 2028Gartner 2026 CIO survey
58%HR adoption for candidate screening. Internal, heavily adopted and regulated, which is the combination that produces a claim rather than a rework.published 2026 AI at work statistics
56 / 51 / 48%adoption in production for customer service, IT operations and marketing, the three leading departments. Two of the three are external-facing.published 2026 AI at work statistics
Some of these sources sell AI spend management, so they have an interest in a high waste figure. Orders of magnitude converge across independent sources, but treat them as such, and never as Versatil measurements.

The number this page cannot compute is the one worth knowing.

Four weeks, one team, one real decision cycle. You leave with your own figure instead of an industry average, and with the twin of your own house.