The number changed with every click, and they had to put their signature on it.
A score that doesn't change behind your back and that you can defend line by line in front of whoever has to sign.
Same input. A different score every time.
To speed up report writing, GLOBIA had started relying on a rough AI-based tool for risk assessment of the companies it analyzed. The problem showed up immediately: entering the exact same data for a company, the AI assigned a score of 68, and trying again ten minutes later, a score of 74.
During meetings with clients, consultants couldn't explain which formulas the final figure came from. And when a company didn't provide all the required documents, the AI generated an estimated score anyway instead of flagging the missing information. Signing an official report based on unstable figures put the firm's professional reputation at risk, and to avoid client objections, consultants were forced to redo the analysis by hand, wiping out any time saved.
On the left, the data entered, always the same. On the right, the score the AI returned at every regeneration. No one had ever built the bridge between the two.
The calculation was automatic. It stayed written in the wrong logic.
The software promised to speed up report writing, but the problem wasn't speed: it was that no fixed formula governed the calculation. The AI generated a plausible score every time, without the same input necessarily producing the same output, and without ever explaining where the figure came from. When documents were missing, instead of flagging it, the model made up a number anyway. A consultant doesn't have the tools to defend in a meeting a grade that not even they can reconstruct: so the report stayed a statistical output instead of becoming a defensible evaluation, and every client objection forced the calculations to be redone by hand in Excel.
An engine that calculates, not guesses.
We built a deterministic evaluation engine where AI never intervenes in the score calculation: only fixed mathematical rules, from the questionnaire to the final number.
"Same data entered: the first time the system returns 68, the second time — ten minutes later, with nothing changed — it returns 74. No one can explain why."
"Same data entered: fixed weighted formulas apply the same calculation parameters. The result is 82 today, 82 tomorrow, 82 always — and every point can be traced back to the item that produced it."
The software applies rigid calculation parameters to generate a 0-100 score: the same starting data always produces the exact same result.
If the analysis lacks essential documents, the system flags "missing data" and blocks the calculation, instead of making up a figure.
Evaluation time drops by 60%, stability is total.
With the new fixed-formula engine, consultants no longer need to double-check calculations by hand in Excel or justify variable grades to clients: the same data always produces the same score, and every objection closes by showing the formula on screen.
From manual Excel checks that dragged on, to a traceable calculation in a few minutes: -60% evaluation time, with an identical score at every regeneration.
From manual Excel checks to an automatic calculation, same data same result.
The same data always generates the same score, no longer an unstable grade.
Resolved by showing the calculation formula on screen, instead of redoing the analysis by hand.
“We used to use an AI tool that changed evaluation grades from one minute to the next, forcing us to redo everything by hand in Excel so we wouldn't look bad in front of clients. With the new fixed-formula engine we have the mathematical certainty that the same data always produces the same grade. We've cut analysis time by 60%, and in meetings we can show the client the exact origin of every single number.”
Head of Due Diligence, Globia
If the score changes behind your back, you don't have a reliable tool: you have a figure no one can defend.
How many hours do you lose re-checking by hand a score that should already stand on its own? The ROIometer puts it in euros.