The short answer
Define AI proficiency as the knowledge and decisions a person needs for a particular job. Write down the task, permitted use of AI, data boundaries, checks and human decision owner before choosing an assessment method.
Replace a vague requirement with an observable one
A job description that asks for 'AI proficiency' without saying what the person will do invites inconsistent judgements. One interviewer may count frequent use of a chatbot, while another expects experience reviewing AI-assisted decisions. Neither can evaluate candidates fairly against an undefined requirement.
Start with an actual task in the role. For example, a talent acquisition partner might use an approved tool to draft outreach. The useful requirement is whether they can select appropriate inputs, check the output for unsupported claims and preserve human responsibility for what is sent. That is more precise than asking candidates to name their favourite model.
Separate a common knowledge baseline from job evidence
Core-6 is the in-market, Sova-delivered assessment of common AI knowledge across six areas. It uses the same generic assessment across roles. A result can add consistent evidence about foundations, risks, data governance, evaluating outputs and human oversight.
A role-specific interview or work sample, if needed, is a separate employer-designed activity. Decide in advance what good evidence looks like, give candidates comparable instructions, consider accessibility and review any use of candidate data. Do not present the separate activity as a Core-6 module or infer job performance from the knowledge score alone.
Check the process as well as the candidate
If AI assists recruitment itself, distinguish that from assessing a candidate's AI knowledge. The UK Information Commissioner's Office has highlighted fairness, transparency and data protection questions for organisations using AI recruitment tools. Review the tool, its data use and the candidate-facing process before relying on its outputs.
Record why the AI requirement matters to the role, the evidence used, who evaluates it and how a candidate can challenge or explain an unexpected result. A hiring decision still belongs to people accountable for the process.
Put it into practice
A simple way to start
- Choose one taskIdentify where approved AI assistance is actually expected in the job.
- Set boundariesSpecify allowed data, required verification and decisions that stay with a person.
- Pick evidenceUse Core-6 for a common knowledge baseline and an employer-designed interview or exercise only where task performance matters.
- Apply consistent criteriaTell candidates what is being assessed, use the same relevant criteria and document how evidence informs the decision.
Sources and further context
- ICO: AI in recruitment UK data protection considerations when using AI tools in recruitment
- ICO: recruitment and selection Guidance on candidate information, fairness and decisions