As of 13 August 2026, AI can only partly analyse an employee skills gap.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open the employee's current role description, competency framework and agreed objectives, then copy the requirements that genuinely apply to the role.
   2. Gather relevant evidence such as work samples, training records, self-assessment and documented observations, and remove unnecessary names, health information and other personal data.
   3. Ask the employee what support, tools, workload or unclear expectations may affect their performance before treating any difference as a skills gap.
   4. Paste the role requirements and anonymised evidence into a chatbot with the supplied prompt, then ask it to separate evidence, assumptions and missing information.
   5. Compare every proposed gap against the original role requirements, work records and the employee's account, correcting or removing claims that cannot be supported.
   6. Discuss the checked analysis with the employee and a suitable HR or management colleague, then agree a documented development plan rather than using the AI output as a performance decision.

   Prompt

   Analyse the possible skills gap for this employee against the role requirements below. Use only the information supplied and do not invent achievements, weaknesses, qualifications or business requirements. Treat the result as a discussion aid, not a performance rating or employment decision.
   
   Role title: [role title]
   Role requirements and current responsibilities:
   [ paste the role description, competency framework or required outcomes ]
   
   Employee evidence, using anonymised information only:
   [ paste relevant work samples, agreed objectives, training records, self-assessment and observed evidence ]
   
   Produce:
   1. A table with each required skill, the available evidence, the current indication, confidence level and evidence missing.
   2. A separate list of likely development needs, clearly labelled as hypotheses where the evidence is incomplete.
   3. A practical development plan with actions, support, resources and a suggested review point, without inventing dates or promising outcomes.
   4. Questions a manager should ask the employee before reaching a conclusion.
   5. Possible alternative explanations, such as unclear expectations, lack of access, workload, insufficient training or reasonable adjustments.
   6. A short section called "Do not conclude from this analysis" listing decisions that require human judgement.
   
   Do not use protected characteristics or irrelevant personal information. Flag any statement that would need checking against an objective record. Keep the tone neutral and suitable for a UK workplace.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

3. Hand it to a person

   The distant third

   A person who owns the outcome does this end to end, worth it when the failure is dear.