Search "Personalvermittlung KI-Fachkräfte" and most of what comes back is a generalist IT staffing agency that added an "AI/ML" filter to its candidate database. Tagging a resume "KI" because it mentions ChatGPT or a Python course is not the same as being able to tell a real AI specialist from someone who has picked up the vocabulary. This is a checklist for what to actually demand before trusting an agency with an AI hire.
The "KI" tag is not vetting
Most database-driven Personalvermittlung agencies work by keyword tagging: a candidate mentions machine learning, TensorFlow or "AI" anywhere on their CV or LinkedIn, and they land in the "KI-Fachkräfte" bucket the agency pitches to clients. This tells you nothing about whether the person can design a retrieval pipeline, debug a production model regression, or reason about cost and latency trade-offs under load. The tag is a search filter, not an assessment.
Question one: who actually evaluates the candidate?
This is the single most important question to ask any agency claiming AI specialization. A generalist recruiter, however experienced in staffing, cannot reliably distinguish a candidate who has actually shipped production ML systems from one who can talk about them fluently. Insist on knowing whether technical evaluation is done by people with hands-on AI/ML shipping experience, whether internal to the agency or a vetted external panel, and what their background actually is.
What a real technical vetting process looks like
- A work sample or take-home exercise close to the actual job, not a generic coding puzzle unrelated to the AI stack in question.
- A live systems discussion: how would you design X, what breaks at scale, how do you handle model drift or data quality issues in production.
- Verification of claimed project ownership, not just "worked on an ML project" but specifically what they owned end to end.
- Sub-specialty differentiation: an ML engineer who builds and deploys models is not the same as an applied researcher, an MLOps engineer, or an LLM/agent engineer, a real assessment tests for the specific sub-specialty the role needs.
Red flags that mean it's a resume database, not a specialist agency
- "AI" is one filter category on their site alongside 40 other unrelated tech tags, with no sub-specialty breakdown.
- The screening call is entirely about availability, salary and soft skills, with no technical component at all.
- They cannot describe, specifically, who evaluates technical depth or what the evaluation actually tests.
- Every candidate they submit "has AI experience" but none can point to a shipped, production system they personally owned.
- Time-to-shortlist for an AI role is identical to their time-to-shortlist for a generic IT role, a sign the process isn't actually different.
A short checklist before you sign
- Ask for their technical assessment rubric in writing.
- Ask who scores it and what that person's AI/ML background is.
- Ask for two reference placements in genuinely senior AI roles, with client contacts.
- Ask how they differentiate ML engineering, MLOps, applied research and LLM/agent engineering in their process.
- Ask what happens, concretely, if the candidate they place turns out not to have the depth they claimed.
