The starting point
The team was already using AI, but chaotically: different people, different prompts, unstable quality and no rules on personal data. They wanted a system, not another generic talk about AI's potential.
What we did
- Ran an AI audit of the processes and identified which repetitive tasks cost the most time.
- Chose scenarios with real value and controlled risk, not "everything at once".
- Built a prompt library and quality standards for intake, job posts and candidate communication.
- Set up an assistant for interview structures and summaries.
- Wrote rules for safe data handling and human review of output.
Why it worked
The tools were built around the team's real tasks, not abstract capabilities. Every scenario had a clear point where a person makes the decision and the final check.
An important boundary
AI made no hiring or assessment decisions. It reduced preparation work — final responsibility stayed with the recruiter.
What the client gained
Concrete AI tools, prompts and rules for the team's real tasks that cut repetitive work and levelled quality — without risks to personal data and without delegating decisions to a machine.