An AI Pilot Where You Set the Success Threshold — Before It Starts You've heard the promises: AI will "streamline processes" and "save your team's time" — no numbers, no answer to what happens if it doesn't work. This pilot is a test on a chosen slice of your data, after which you know whether the automation reaches a threshold agreed in advance.
Ready to publish without client-provided data. --- # An AI Pilot Where You Set the Success Threshold — Before It Starts
You've heard the promises: AI will "streamline processes" and "save your team's time" — no numbers, no answer to what happens if it doesn't work. This pilot is a test on a chosen slice of your data, after which you know whether the automation reaches a threshold agreed in advance.
A vendor shows up with a ready-made implementation offer — no test, no intermediate stage, straight to a full project and a full invoice. If the solution doesn't work on your real, non-standard data, you find out only after paying for everything.
That's a risk you don't have to take. Before anyone proposes full implementation, the hypothesis can be tested on a limited sample — with a clear success threshold set upfront, not after the fact.
We run a time- and budget-limited test on an anonymised sample of your data. The result is unambiguous: threshold met or not — not a general impression, a concrete answer.
Before writing a single line of prompt or integration code, we jointly set a success threshold — a specific number, such as the share of correctly classified cases. Every model decision during the pilot is logged, and results are reviewed by a human before we call the pilot complete. Data is processed in the EU or locally on your infrastructure, depending on what we agree.
What we don't do: we never run a pilot on synthetic data, because that result proves nothing. We work on a real, anonymised sample.
This is a new service line, and we don't yet have completed implementations to show — we say that plainly rather than inventing references. Our first pilots run at a preferential rate in exchange for the right to describe the outcome as a reference case.
The pilot covers one process, ideally already measured or described in enough detail to set a realistic threshold.
Price depends on the process's complexity, the type of input data, and whether integration with your system is needed already at the test stage. The pilot's price is fixed and capped — it doesn't grow mid-test, because the scope is closed from the start.
Threshold set before the start. In writing, before any work begins — your decision, not ours.
No threshold, no implementation. If the result doesn't meet the threshold, we don't propose implementation and you pay nothing beyond the pilot itself.
Code in your repository. Code, prompts and documentation come to you regardless of the pilot's outcome.
The pilot is run by the same small team responsible for the result's quality, so we run a limited number of pilots in parallel — the start date is set after the threshold workshop.
Send us a description of the process you want to test — ideally with the result of an earlier measurement. We'll reply within a few business days with a proposed threshold and a firm price.
Every month without a pilot is a month of deciding on automation without checking whether it even works on your real, non-standard data.
Success threshold set in writing before the start, on your terms · test on an anonymised sample of real data · human in the loop and a decision log · no threshold means no implementation · code and prompts land in your repository regardless of the outcome.
Yes, it's your decision — we help define it realistically, e.g. based on the result of a process measurement, so it's measurable and achievable.
No — we work on an anonymised sample sufficient for a credible test, without you handing over your full dataset.
We delete the test data after the pilot ends, unless agreed otherwise in writing. Code and prompts stay in your repository regardless of what happens to the data.
Yes, at any point — billing is proportional to the stage completed, under the terms agreed before the start.
Basic integration is sometimes needed for the test itself, but full production integration is priced and delivered at the implementation stage, if the pilot clears the threshold.