References for the AI yes are not a call log.
They are a yes-test.
What did the bearer decide, stop, and prove before?
Run references as an HR checklist and you check attendance. Run them as a yes-test and you check judgment.
I run retained AI Executive Search mandates from Hamburg for DACH seats. This page is the yes-test: not reference-architecture and not the HR reference farm.
They prove prior Träger behavior.
Not 'whether a model went live.'
Not 'whether the title matched.'
But: Which yes was carried under pressure? Which stop without theatre? Which proof held?
References for the AI yes ask for decisions: not project names.
Not reference-architecture as a process spine.
Not an HR checklist of 'three calls, two notes.'
Not a find-how-to and not a credentials farm.
A one-line pointer to the architecture page is fine. The spine here is references as yes-test.
Cuts: What was killed while the line screamed?
Ownership: Which calls did the bearer carry alone?
Proof: What did a board accept as evidence: not as story?
If all you get is 'team player' and 'strong AI passion,' the yes-test failed.
Because CVs sell titles.
Because interviews sell hope.
Because only references as yes-test show whether judgment repeated.
If you run references after the offer, you already gave the yes.
If the AI yes is real, retain the AI Executive Search mandate with references as yes-test: not as formality.
If you only tick boxes, you are still searching blind.
No. Architecture builds process. This tests yes-content.
No. Written yes-questions to real decision-makers: not soft endorsements.
Beyond Chiefs with whoever owns the yes: not HR alone.
