What AI really delivers — we measured it
"AI saves 40 % of your time." Everyone knows figures like that — but where do they come from? We accompanied a real rollout for 30 days, classified every single chat and worked out use case by use case how much work AI actually saves. No estimate, no slide from an analyst report — measured. We take you through the study step by step.
Step 1 — we analysed it
Mid-sized wholesale company · November 2025 · anonymised
An entirely normal month at an entirely normal company. Not a lighthouse project, but the working day of 170 people across every department.
Step 2 — how the study was set up
Instead of one blanket productivity figure, we assigned every chat to a category, rated its complexity and its success — and linked it to a use-case catalogue that records a conservative time saving for each task. Only that produces a real ROI.
Step 3 — there are plenty of use cases
Almost a quarter of all requests are complex tasks — the platform is not used for playing around but for real problem solving.
Step 4 — what the AI is actually used for
learning phase — not in the ROI
risk reduction — not in the ROI
Every category has its own story. In the use-case map you can zoom into each department — with real tasks and example prompts.
Step 5 — three kinds of value
We deliberately separated the benefit into three pots — and only converted the first two into euros:
- Time saved — AI speeds up repetitive tasks.
- Enablement — AI makes possible what would otherwise not be done at all.
- Risk reduction — AI supports critical checks (sanctions, customs, contracts).
From pure time alone, this is what adds up in one month:
| Use case | Usages | Saving each | Total |
|---|---|---|---|
| Writing & emails | 1,166 | 8 min | 155.5 h |
| Technical research | 398 | 12 min | 79.6 h |
| IT support | 327 | 10 min | 54.5 h |
| Strategy & planning | 235 | 12 min | 47.0 h |
| Coding | 185 | 15 min | 46.3 h |
| Excel & data | 215 | 10 min | 35.8 h |
| Translation (app) | 403 | 5 min | 33.6 h |
| Workflows | 230 | 8 min | 30.7 h |
| Further use cases | … | … | ~67 h |
| Total time saved · 1 month | ~550 h | ||
Step 6 — what does that deliver per person?
1 pilot user · 1 month25 chats × 7.8 min
= 3.3 h × €50/h
images + BI research
+ translation + meetings
~1.5 compliance checks
sanctions · terms · customs
A single user delivers around €2,800 a year — and that is before risk reduction, quality and satisfaction are counted at all.
What we deliberately did NOT count
The calculation is intentionally cautious. These real effects are not included in the value:
- General knowledge (11 % of chats) — pure AI learning phase
- Risk reduction — compliance breaches avoided, valued at €0
- Quality gains — better texts, fewer errors
- Employee satisfaction — hard to express in euros
- Costs avoided — DeepL subscriptions, stock photos, external consultants
The value stated is therefore a lower bound, not a best case.
What this means for you
The lever transfers: the value arises in the breadth — across every department, not in a single lighthouse use case. That is exactly what the use-case map shows: where AI really lands in everyday work.
Start with AI in your company
Together we find the right use cases, connect your systems, and bring AI into daily work in line with your business.