Roughly one in three AI projects is abandoned after the pilot. Yours doesn't have to be. Describe one of your teams: see the time AI could hand back to it every year, valued in euros. Clearly, and right away.
Valued at the team's cost, on a deliberately conservative assumption.
Agent, automation or copilot: AI prepares the work, your teams keep the final say.
“Time handed back” = time currently absorbed by repetitive tasks, which AI frees up and the team reallocates to higher-value work. Of the €72,000 these tasks already cost the team every year, we prudently count only a share (never 100%). It's capacity handed back, valued at its cost, not cash in the bank.
Indicative estimate. It doesn't replace an audit: the real numbers depend on your processes.
1 Gartner, 2024: generative-AI projects abandoned after proof of concept. 2 Controlled studies: NBER / Brynjolfsson (+14%), BCG × Harvard, GitHub (up to +55% on targeted tasks). 3 Market benchmarks: public price lists of specialised French AI agencies, 2025–2026. 4 McKinsey, 2023: economic potential of GenAI.
Your on-screen estimate, broken down and turned into an action plan. We'll email it to you:
The calculator works from the value of time, not from magical thinking. We start from the team's annual fully loaded cost (salaries + employer costs), multiplied by the share of time spent on repetitive, document-heavy tasks. We then apply a deliberately conservative recovery rate, specific to the role (around 18 to 22% of the targeted time, never 100%), to get the “time handed back” per year, valued in euros.
Time handed back / yr = annual loaded cost × share of automatable tasks × recovery rate. The full ROI — (gains − costs) ÷ costs — is then worked out on a real quote, after scoping.
And the investment? The tool doesn't estimate it: it depends too much on scope, data and the level of customisation to be captured by a formula. Every project is quoted after a scoping call, on its real scope.
For price benchmarks by project type, read our guide: how much an AI project costs for an SME in 2026.
It's the time currently absorbed by repetitive tasks that AI can free up and the team can reallocate to higher-value work, valued at the team's fully loaded cost. It's not cash in the bank, but capacity handed back. We prudently count only a share of it, never 100%.
We multiply the team's annual fully loaded cost by the share of time spent on automatable tasks, then apply a conservative recovery rate specific to the role (18 to 22%, never 100%). The result is the time handed back per year, valued in euros. The tool doesn't estimate the investment: every project is quoted after scoping.
It varies a lot: it can start quite low with subscriptions and light configuration, and go quite far with custom development. It all depends on the ask, the scope and the level of integration. Details and benchmarks in our guide on AI project costs.
It's an order of magnitude, not a quote. The calculation rests on a deliberately conservative assumption of recoverable time (18 to 22% of the targeted time, below the gains measured by public studies), and the real cost of a project varies widely with the ask and the level of customisation. It doesn't replace an audit: the real numbers depend on your processes. It's there to frame the conversation before going further.
A two-to-three-week audit turns this estimate into a costed plan, team by team. Which one do we start with?
Book a call