Decide with evidence
Distinguish what the AI claims from what an expert has actually checked on site or against the relevant object.
AI tools · Applications · Robotics
Does your AI product
Does your AI product
deliver on its promises?
Your AI produced a result. A qualified domain expert checks it on site or against the real object, corrects errors and documents the evidence. You know what you can rely on.
Automotive01 / 11
Human expertise
behind AI.
Illustrative visuals
What you receive
Facts for a confident decision.
AI publisher, operator or holder of an AI report: compare its factual claims with reality before relying on them.
Prioritise improvements
Identify errors and omissions, understand their severity and obtain corrections linked to evidence.
Check progress
A corrected result can be rechecked. A reference and public QR record identify the agreed report version.
Our approach on film
Behind every AI product, a real-world context.
Three short films about the role of human expertise in AI-powered applications.
Observe a property. Question the AI’s analysis.
Three professional worlds
Playback starts automatically with sound off. You can enable it in the player; after all three films, sound is turned off for the rest of the loop.
Promotional films in French. Illustrative scenarios.
Read a summary of the films
Property. An AI-powered property tool is compared with real-world needs: space planning, reading a floor plan, the accuracy of a sales description and the relevance of a rental management assistant’s response. A domain expert questions the analysis and documents its limitations. Scenes and interfaces are illustrative.
Automotive & trackdays. Two laps on a circuit illustrate the evaluation of AI-powered equipment. Simulated speed, brake and accelerator charts, not measured from the footage, and a trajectory diagram raise questions about interpretation and grip. Founder Jean-David Jaïs presents the role of domain expertise in evaluating these analyses and their limitations.
Restaurants & hospitality. An illustrative restaurant scene raises questions about AI-generated menu descriptions, service planning, reservations and missing information. The professional and interfaces shown are fictional; the film presents no measured test result.
Hu-score · checked against reality
A clear view of your AI result’s reliability.
Four levels, awarded after a field audit: verified facts, omissions, error severity and traceable evidence. Each score applies to a defined result, version and scope.
Insufficient evidence means no overall score. The Hu-score is not a certification.
What has your AI reported?
Describe the result to check, the site or object concerned and the intended use. We agree the expertise, evidence and quote before the audit.
FAQ
Your questions, answered.
What products do you evaluate?
We audit factual results produced by AI tools, AI-enabled applications and physical AI through robotics, by comparing them with the real site or object. Explore our fields of expertise.
What do I receive?
A report identifying verified facts, errors, omissions, corrections and traceable evidence. A Hu-score is awarded only when the field audit supports it; a public QR record identifies the agreed report version without exposing private evidence. See our method.
How do I request an evaluation?
Start with the five-question pre-diagnostic, then describe the result to check. We confirm the qualified expert, site access, scope, schedule and quote. Tell us about your project.