Why your buyer asks
Large customers have to be able to explain which AI they use and what risks come with it. That's why the question lands with you, often in the middle of a deal.
“Yes” is no longer enough
More and more buyers ask for evidence: documents, settings and test results. Marketing copy doesn't count.
AI doesn't always answer the same
One successful demo says little. Repeated tests give more evidence about behaviour within the tested scope. They are not a guarantee of future behaviour.
Better you than your customer
If your buyer finds the problem, it costs trust and sometimes the deal. If you find it first, it's an improvement.
What you get
Examples of claims we test
We only test what you claim. A few common examples:
Data separation
“The assistant never shows another customer's data.”
Roles and permissions
“The AI sees and does no more than the user is allowed to.”
Instructions in documents
“The assistant doesn't follow instructions from uploaded files or emails.”
Source references
“Every answer cites a source the user can check.”
How it works
- 1
20-minute call
Is there a review, which AI claims matter, and when does it need to be ready?
- 2
NDA and Rules of Engagement
Written agreements on test environment, time window, contact person and stop procedure.
- 3
Claims, evidence and tests
We link each claim to your documentation and test it repeatedly with fictional data.
- 4
Report and retest
You see the outcome first. After a fix we test once more. You decide what to share.
Four outcomes, no score
No number from 0 to 100 and no seal. Per claim one of these four outcomes, with the evidence attached.
Frequently asked questions
Is this a pentest?
No. We test the specific claims you make about your AI features, with repeated scenarios. A full penetration test or red team is out of scope.
Is this an EU AI Act assessment?
No. We don't provide a conformity assessment, certification or legal advice. We do provide evidence and test results you can use in such a process.
What if a claim isn't confirmed?
You hear it first, before anything goes to your buyer. You can fix it; one retest is included. You decide what to share.
Do you test in production?
Never. Only in a test or staging environment you designate, with fictional data, within a fixed time window and with written authorisation.
How long does it take?
5–10 working days after complete intake and test access. For complex assessments, such as agents that take actions or several products, we quote from €5,000.
Discuss your AI claims
A short 20-minute call: is there a review, which claims matter and what's feasible for your deadline. Then you know what it costs.
- No confidential information needed for the first call
- Testing only with written authorisation, never in production
- Delivery 5–10 working days after intake and test access
Rather email first? hello@holdvero.com
Other services
Questionnaire Rescue
Security, privacy or AI questionnaire up to 150 questions
€695
More about Questionnaire Rescue →Complex questionnaire
Portals, large reviews and tenders
From €1,195
More about Complex questionnaire →Evidence Maintenance
Keep approved answers and evidence current
Custom
More about Evidence Maintenance →Continuous Assurance
Periodic reassessment of agreed AI claims
Custom
More about Continuous Assurance →