Preparing for a performance review with AI: safely

Preparing for a performance review with AI can quickly give you more overview once the invitation is already in your calendar. Maybe the meeting is in a week and you no longer remember exactly what you achieved in March.

Or you are a manager and you have eight meetings back to back. With the right approach, you can prepare clearly in 30 minutes, without pasting names or sensitive data into a chatbot.

Can you prepare for a performance review with AI?

Yes, you can prepare for a performance review with AI as a tool for structure and wording. Only do this without entering data that can be traced back to a person.

AI is especially useful if you already have loose notes. Think of results, difficult moments, questions and wishes for the year ahead. The chatbot can turn that into a clear order.

AI can also help you make a difficult sentence sound calmer. For example, if you want to give feedback or ask something about development, workload or career. You then decide whether the text is right and suits you.

This is what AI does well:

• Organising your own notes into results, lessons and wishes.
• Creating a short summary of points you provide.
• Suggesting questions you can ask in the meeting.
• Thinking of counterarguments so you are better prepared.
• Practising a conversation with a fictional role.

This is what AI does not do well:

• Assessing your work fairly without full context.
• Knowing facts you have not given it.
• Deciding whether you deserve a good review.
• Knowing how your manager will really react.

So use AI as a writing partner, not as an assessor. The best preparation starts with your own facts. Then AI helps turn them into a clear story.

How do you prepare for a performance review with AI as an employee?

As an employee, start with rough points from your own work. Then let AI organise them into results, lessons and wishes.

Many people start too late with their preparation. Then you forget small successes, improvements and agreements from the year. AI helps most when you first collect the basics yourself.

Work in four steps:

1. Gather notes from your calendar, emails and your own task list.
2. Remove names, client details and other identifying information.
3. Use a clear prompt with what you want back.
4. Check the result and rewrite it in your own words.

A prompt is the instruction you give AI. If you want to understand the basics first, also read what a prompt is.

Be careful not to ask AI to invent results. Ask for ordering, not imagination. If information is missing, AI should ask you for it.

You can, for example, divide your preparation into three blocks. First, what went well. Then, what you learned. Finally, what you need in order to keep growing.

Good preparation does not have to be long. A few clear points are better than a long text that does not sound natural. Only include what you can explain.

Use the output as a draft. Then add what is missing. Remove what is not right. And make the sentences shorter if they do not sound like your own way of speaking.

How do you prepare for a performance review with AI as a manager?

As a manager, use AI for the structure of the meeting, open questions and phrasing difficult feedback. Do not enter case file details of an employee into a regular chatbot.

AI can help you build the meeting more calmly and clearly. For example, with an opening, questions about collaboration and space for development. AI can also make feedback sound less harsh or more specific.

When giving feedback, use a simple structure:

1. Situation: when or in what context did something happen?
2. Behaviour: what did someone actually do?
3. Effect: what was the result for the work, team or customer?
4. Follow-up: what do you want to agree for the coming period?

This structure helps you stick to the facts. Do not just say that someone needs to communicate better. Explain the situation it happened in and what effect it had.

Never let AI write an opinion that you cannot support yourself. You lead the conversation. You weigh the context. AI only helps you organise your words.

For entrepreneurs and managers, clear internal policy is important. Record which tools staff may use and which data must never go into AI. A starting point for that is setting up an AI policy.

At Dutch employers, the works council may have a role in systems that process staff data. Be careful with that and check internally which rules apply.

What data must you not put into ChatGPT or Copilot?

Do not put data into a regular chatbot that could identify a person. Special categories of data, such as health information, definitely do not belong there.

The Dutch Data Protection Authority warns that employees can cause a data breach if they enter personal data into AI chatbots such as ChatGPT or Copilot, because the provider may be able to see the data, according to Mediafederatie. That does not mean every use of AI is forbidden. It does mean you should carefully consider the risks, your employer’s permission and the legal basis on which personal data is processed.

There is a difference between a private account and a business account with agreements from your employer. With a private account, you often have a less clear idea of what happens to the information. With a business account, clear rules may have been agreed.

So always ask what your employer allows. Also read the rules for your privacy before you process sensitive work issues in AI.

Okay to enter	Better not	Never
Your own results without names	Results with identifying details	Assessments of colleagues
General project description	Project names if they make people identifiable	Secret or sensitive project information
Roles such as [colleague A]	Job titles if only one person fits	Names of colleagues or clients
General wish for development	Details about private problems	Medical information or health data
A general question about pay	Your own salary details in a regular chatbot	Salary details of colleagues

ChatGPT, Copilot and Gemini work well in Dutch. Still, the rule remains the same. Only give the tool information that you are allowed to share safely.

In Belgium, this meeting is often called an evaluation interview. The GDPR privacy rules apply in the Netherlands and Belgium. There too, careful handling of personal data is important.

Can you practise the conversation with AI?

Yes, you can practise the conversation with AI. Let the chatbot play the role of manager or employee with a fictional profile.

That way you can practise answering, staying calm and asking follow-up questions. That can feel reassuring if you want to discuss a difficult topic. Think of career, workload, collaboration or salary.

Make it clear that it is training. AI does not predict what the other person will really say. It only provides a practice situation.

After the exercise, you can ask for feedback. For example, whether your answers were clear, concrete and calm. Use that feedback as a mirror, not as the truth.

As a manager, you can also practise giving difficult feedback. Use a fictional employee for that. Do not paste a real case file into the chatbot.

What does the law say about AI in assessments?

The GDPR in principle forbids decisions about people that are made solely by automated means and have legal or similarly significant effects. That is set out in article 22 of the GDPR.

According to the Dutch Data Protection Authority, this concerns fully automated individual decision-making, including profiling, with legal or similarly significant effects, as described in the document on supervision of AI and algorithms. That does not mean every use of AI around meetings is forbidden. It does mean that an assessment with major consequences may not simply be determined by AI alone.

For you as an employee, this changes little if you only use AI to prepare your own meeting. You then use AI as a tool for structure. You do not let the tool make a decision about someone.

For employers, it is different. AI used for recruitment or employee assessment falls under high risk in Annex III of the AI Act. The AI Act rules are being phased in gradually. Employers who want to use such tools must therefore carefully check which rules, controls and internal agreements apply.

That does not mean employers can do everything until then. Privacy, good employment practice and internal agreements remain important. So do not use AI as a replacement for human judgement.

Practical example: preparing for a performance review with AI

This example shows how you can work safely and practically. It is a fictional example, not an AIzine test and not a conversation with a real person.

Sanne is a project coordinator at a municipal organisation in the Utrecht region. In two weeks she has her annual review. She does not want to be looking for what she did at the last minute.

She takes this approach:

1. She makes a list of eight loose points from her calendar and emails.
2. She removes the names of colleagues and clients.
3. She pastes the anonymous notes into ChatGPT or Copilot with a clear prompt.
4. She gets an organised overview with three results, two lessons learned, a wish for a training course and five questions for her manager.
5. She checks each point, adds figures AI does not know and removes what is not right.

The result is not a final report. It is a usable draft. Sanne then adapts it in her own words.

That last part matters. If you copy the text word for word, it can sound distant or too polished. In the meeting, you need to be able to explain it yourself.

Common mistakes when preparing for a performance review with AI

Most mistakes happen because people share too much information or trust AI too much. You can avoid that with simple rules.

These are the mistakes that often go wrong:

• Pasting names, salary or assessments of colleagues into the chatbot. Use role labels such as [colleague A] and work with an approved business account.
• Letting AI invent results. Always add: do not invent anything. Check every figure and watch out for AI hallucinations.
• Copying the text word for word. Rewrite the output in your own words, otherwise you will not sound like yourself.
• As a manager, letting AI determine the judgement. You assess, AI only helps with wording.
• Not checking what the employer’s AI policy says. Ask before you start.

A good rule of thumb is simple. If you would not casually share the information with an external party, do not paste it into a regular chatbot either.

Frequently asked questions

Can I paste my performance review report into ChatGPT?

Better not, unless the report is properly anonymised and your employer allows it. Remove names, identifiable situations, salary details and medical information.

Can my boss have AI assess me?

A fully automated decision with legal or similarly significant effects is in principle forbidden under article 22 of the GDPR. AI may support the process, but a human must make the decision.

Which AI is best for this purpose?

Any common assistant is fine for organising, wording and practising. Preferably choose the business account offered by your employer.

Can AI improve my goals or SMART wording?

Yes, give your rough goal and ask for a measurable version. Then check whether the goal really fits your work and agreements.

Do I have to say that I used AI?

Check this in your employer’s AI policy. Being honest is wise, especially if AI helped with a text that you share.

Copy the prompt, add your own notes today and remove identifying data. If you want to prepare for your performance review with AI within clear workplace rules, read more about setting up an AI policy.

• ChatGPT
• Copilot
• Privacy
• Prompts