AI Employee Coaching Assessment Checklist

AI coaching tools may allow employees to practise workplace conversations and receive automated feedback. Use this checklist before introducing a platform that records, transcribes, scores or analyses employee interactions.

Step 1 — Identify the use case

• □ Supplier
• □ Product
• □ Scenario
• □ Business purpose
• □ Business owner
• □ HR owner
• □ Technical owner
• □ Intended users
• □ Voluntary or mandatory
• □ Pilot or production

Step 2 — Define the purpose

• □ Private practice
• □ Developmental coaching
• □ Training completion
• □ Skills assessment
• □ Recruitment
• □ Performance management
• □ Promotion
• □ Compliance evidence
• □ Other (specify)

Do not describe the system as coaching if the results will be used for formal employment decisions without saying so clearly.

Step 3 — Map the data

Does the system capture:

• □ Name
• □ Employee ID
• □ Voice
• □ Video
• □ Transcript
• □ Score
• □ Feedback
• □ Repeated attempts
• □ Session duration
• □ Facial data
• □ Device data
• □ Manager comments

For each item checked, record: purpose · location · retention period · access controls · deletion method.

Step 4 — Review visibility

Who can see:

• □ Session completion
• □ Transcript
• □ Recording
• □ Feedback
• □ Score
• □ Number of attempts
• □ Historic trends
• □ Team comparison
• □ Individual ranking

Prefer private practice and aggregate reporting where possible.

Step 5 — Review the rubric

• □ Criteria documented
• □ Business owner assigned
• □ HR review completed
• □ Employee-facing explanation available
• □ Tested across different communication styles
• □ Tested across accents
• □ Tested across supported languages
• □ Accessibility reviewed
• □ Weighting understood
• □ Pass threshold justified
• □ Human challenge process available

Step 6 — Review fairness and accessibility

• □ Alternative participation method available
• □ Reasonable adjustments defined
• □ Speech impairments considered
• □ Hearing impairments considered
• □ Neurodivergence considered
• □ First-language differences considered
• □ Accent differences tested
• □ Assistive technology supported
• □ Employees can request human coaching

Step 7 — Review data protection

• □ Privacy notice provided
• □ Lawful basis documented
• □ Necessity assessed
• □ Proportionality assessed
• □ DPIA completed where appropriate
• □ Retention limited
• □ Access restricted
• □ Employee rights supported
• □ Subprocessors reviewed
• □ Customer-data training setting understood

Step 8 — Review employment use

Will the score affect:

• □ Appraisal
• □ Probation
• □ Promotion
• □ Pay
• □ Training access
• □ Shift allocation
• □ Disciplinary action
• □ Dismissal
• □ Recruitment

Where yes, confirm:

• □ Human review is meaningful and independent
• □ Employee can challenge the result
• □ Independent evidence is considered alongside the score
• □ Legal and HR review completed

Step 9 — Review security

• □ Single sign-on
• □ Multi-factor authentication
• □ Role-based permissions
• □ Encryption in transit and at rest
• □ Audit logging
• □ Secure deletion
• □ Export controls
• □ Tenant separation
• □ Incident notification terms
• □ Contract reviewed

Step 10 — Run a safe pilot

• □ Low-risk scenario selected
• □ Volunteers used
• □ Scores not linked to HR decisions
• □ Private practice enabled
• □ Short retention period set
• □ Employee feedback collected
• □ Accuracy tested
• □ Accessibility tested
• □ Human coaching retained
• □ Stop criteria agreed

Step 11 — Make a decision

• □ Approved for private practice
• □ Approved for limited pilot
• □ Approved with restrictions
• □ Further evidence required
• □ DPIA required
• □ Equality review required
• □ Legal review required
• □ Not approved

Record: decision maker · date · restrictions · corrective actions · action owner · next review date.