The HR director’s guide to AI performance management

Evaluate AI assistance for performance reviews with a practical framework for context, human oversight, data handling and supplier questions. This practical guide to AI performance management brings together a step-by-step approach, illustrative examples, a reusable worksheet and answers to common questions. Start with the section closest to your current challenge, then use the working session to turn the guidance into a clear next action.

Begin with a specific use case

Evaluate AI against a real task, such as preparing a manager briefing from authorised performance notes. Define what a useful output would contain and what it must not infer.

Avoid starting with a broad promise to automate performance management. That framing hides the different risks and responsibilities of summarising notes, coaching a manager and making an employment decision.

Map the information that enters the system

List the information needed for the task and where it comes from. Distinguish employee-provided reflection, observed feedback and a manager’s interpretation.

• Identify the source and owner of each input.
• Confirm that the user has permission to access it.
• Define which information must be excluded.
• Ask how data is retained, deleted and used by model providers.
• Check how access changes when a person changes role.

Keep security questions tied to evidence. A policy statement is different from a tested technical control.

Keep source evidence visible

A generated summary should help a manager navigate the evidence. It should not create a new layer of facts that cannot be traced.

Test whether the output distinguishes a goal from an achievement, a single comment from a pattern and an observation from an interpretation. Ask the manager to verify every material statement against the original context.

When inputs disagree, the output should preserve the difference for discussion. Averaging conflicting accounts can conceal the information the manager most needs to understand.

Define the human review

Name the person responsible for checking the output and explain what they must check. Give employees a route to correct inaccurate records.

• Check factual accuracy and missing context.
• Remove unsupported assumptions about motivation or personality.
• Confirm that the language is specific and useful.
• Check that authorised information stayed within the right audience.
• Record the manager’s final reasoning where your process requires it.

AI can support preparation; accountability for a review and its consequences stays with the organisation.

Evaluate with realistic examples

Build a small test set that reflects your actual work. Include a changed objective, a new employee with limited history, conflicting project feedback and a matrix reporting relationship. Use fictionalised or approved test information.

Ask different reviewers to assess whether the output is accurate, helpful and appropriately cautious. Record failures, not only successful examples. Confirm how the supplier responds when behaviour changes after an update.

Ask procurement questions that matter

Request evidence on data handling, access controls, subprocessors, retention, model changes and incident response. Confirm which capabilities are available now and which are planned.

Use the software RFP template to compare written answers. Ask the supplier to demonstrate a representative scenario and retain a record of any unresolved question.

Roll out to a small group first. Provide clear instructions, collect feedback and review whether the assistance improves the quality of manager preparation.

Put the guide into practice

Set aside a working session with the people who own this process and one or two people who experience it. Use a fictional or appropriately authorised case, so the discussion can be specific without sharing unnecessary personal information. The purpose is to leave with a usable decision or document, not just agreement that the topic matters.

Prepare the case

Create a fictional review case containing an original goal, a changed deadline, positive feedback and an unresolved disagreement. Define the preparation task precisely, such as locating relevant evidence or suggesting discussion questions. Do not ask the assistant to decide a rating or infer the employee’s motivation from the notes.

Write the starting assumptions down before discussing solutions. If the group disagrees on what happened, identify the information needed to resolve that difference rather than building a plan on an untested story.

Work through the decision

Compare the generated summary with each source. Check whether achievements, plans and interpretations remain distinct. Ask the reviewer to identify an unsupported conclusion and explain the correction. A source link should lead to context the reviewer can access, not merely decorate a statement that cannot be verified.

Ask each participant to explain the proposed decision in their own words. Differences in interpretation often reveal an unclear criterion, a missing responsibility or an instruction that will be difficult to follow.

Test an exception

Introduce limited history for a new starter or conflicting observations from project managers. Check that the assistant preserves uncertainty rather than filling gaps or averaging incompatible accounts. Test access boundaries when the reviewer changes role. Information should remain controlled by the application, not by a prompt requesting discretion.

Record what changes in this situation and what remains the same. An exception should lead to a clear next step, with an owner, rather than an informal workaround that nobody can explain later.

Agree the handoff

Give the manager a preparation record that distinguishes verified facts, questions and unresolved issues. Confirm the employee’s route to add context or correct a record. The final assessment and reasoning belong to the responsible people; generated wording should not silently become the organisation’s decision.

Finish by confirming the owner, the next action and the date when the result will be reviewed. Give the person receiving the work enough context to continue without repeating the whole discussion.

Frequently asked questions

Can AI write an employee performance review?

It may help prepare a draft from approved information, but the organisation must decide whether that use is appropriate and how it is reviewed. A manager needs to verify facts, context, fairness and relevance before using the wording. AI output is not independent evidence of performance. Keep accountability for the assessment, the conversation and any consequential decision with authorised people.

What information should be excluded from AI review preparation?

Use only information approved for the task and the tool. Exclude unnecessary personal details, confidential support information and records outside the reviewer’s access. The exact boundary depends on the organisation’s reviewed data arrangements. Removing names is not enough if the context identifies an employee. Map inputs explicitly and test permissions rather than relying only on a general instruction to be careful.

How do we know the assistance is useful?

Compare the complete preparation process with and without the assistant using representative cases. Examine factual accuracy, missing context, source traceability, review effort and the quality of questions managers bring to the conversation. Include difficult cases and record failures. A faster first draft or a positive demonstration does not establish that the final review is better or easier to verify.

Review the first cycle

Check whether managers notice and correct material errors under ordinary workload. Ask employees whether the conversation reflects their actual work and relevant context. If the assistance encourages overconfidence or takes too long to verify, narrow the task or stop it. Re-evaluate after significant changes to the model, inputs or review process.

Keep a brief record of what was tried, what participants found useful and what needs to change. Compare the result with the original problem rather than judging success only by completion. If the process created extra work without improving clarity, quality or support, simplify it and test again. Share the agreed change with the people who will use it, and name the person responsible for keeping the guidance current.

About New Dynamics

New Dynamics connects goals, feedback, recognition and reviews around the way organisations work. This guide is published by the New Dynamics Editorial Team as part of our practical library for HR leaders, managers and People teams.

Use the examples and worksheets to structure your own discussions and adapt them to your organisation. Illustrative scenarios are not customer case studies. Policy and employment guidance needs appropriate local review before adoption.

For questions about this guide, corrections or a conversation about your performance management process, email contact@new-dynamics.com. Explore the complete guide library for related resources.