Business
Intermediate
16 min
Bias, fairness and the decisions AI must not influence
The problem
You use AI to help screen 300 CVs. Each summary looks fair and each shortlisting decision is defensible. Across 300, candidates from certain colleges, with certain names, or with employment gaps have been systematically ranked lower.
The workflow
1Write the criteria first
2Extract facts only
3Shortlist yourself
4Apply criteria consistently
5Check the aggregate
6Investigate patterns
7Test language parity
★Record
What you need
Process: 300 CVs arrive for an accounts executive role. AI summarises each into a standard format.
The tool
AI extracting stated facts into a table
The prompt
Subscribers. The full prompt is part of the SMEMinds AI Playbook. See plans.
What comes back
1. WHERE AI IS INFLUENCING A DECISION — the summarisation step.
Verify before you use it
Human review required. Check every line below before this leaves your screen.
- Write the criteria before the next round. Their absence is the largest problem and everything else depends on it.
- The summarisation point is the finding most people resist. Reading 300 summaries is not reading 300 CVs.
- Strip name, address and photograph before the AI sees anything. It is a practical step and it removes the most direct proxies.
- Run the aggregate check after the next round even if you change nothing else. You cannot manage what you have never measured.