Day 5: Draft a Performance Review

The concept

Performance reviews should create clarity. Too often, they create confusion.

Managers sit down with scattered notes, old objectives, half-remembered examples, and a form that asks for balanced commentary. They try to be fair, but the final review may still be vague, inconsistent, or too polished to be useful. Employees leave the conversation unsure what they should keep doing, what they should change, and how their manager actually sees their performance.

AI can help managers turn rough notes into a structured review. But HR must set the standard: evidence first, fair language, clear expectations, and no artificial softening.

Plain English

A good performance review says what happened, why it mattered, what needs to change, and what support comes next.

Start with evidence, not adjectives

The biggest weakness in many reviews is adjective-heavy language.

"Great attitude." "Not strategic enough." "Needs more confidence." "Excellent team player." These statements may contain a real judgement, but they do not give the employee enough information to act.

Evidence-based feedback describes observable behaviour and impact.

Instead of:

You need to be more proactive.

Write:

In three project meetings this quarter, risks were raised only after deadlines had already moved. Next quarter, I need you to flag delivery risks as soon as you identify them, with a proposed mitigation.

Instead of:

You are a strong collaborator.

Write:

You helped the finance and operations teams resolve the reporting handoff by setting a weekly check-in, clarifying owners, and reducing repeated data requests.

Specificity makes praise more meaningful and difficult feedback more useful.

AI can structure, but it cannot own the judgement

AI is good at organising rough notes into a coherent review. It can group examples, clarify language, identify missing evidence, and flag risky phrasing. It cannot decide what the rating should be. It cannot know whether the manager has left out important context. It cannot know whether a performance issue was caused by unclear expectations, workload, team instability, or genuine skill gaps.

Use AI as a drafting partner, not as the performance authority.

Before accepting the output, ask:

• Is the assessment accurate?
• Is the tone honest?
• Are the examples real?
• Has the model softened serious feedback?
• Has it inflated ordinary work into exceptional performance?
• Are there claims without evidence?
• Would the employee understand what to do next?

The manager remains accountable for the review.

Preserve the manager's voice

Performance review writing often becomes strange because managers try to sound like HR. The result is stiff language that nobody would use in conversation.

Good HR support does not erase the manager's voice. It helps the manager be clearer, fairer, and more specific.

If a manager would naturally say:

You have become the person the team relies on when client deadlines move.

Do not turn it into:

The employee demonstrates cross-functional reliability in dynamic delivery environments.

The second version sounds official, but it says less.

Encourage managers to write in plain language. The review should be professional, but it should still sound like a real person prepared it carefully.

Separate achievements from development areas

Many reviews blur positive feedback and development feedback together. That makes both weaker.

Achievements should explain:

• what the employee did well
• what evidence supports it
• what impact it had
• why the behaviour should continue

Development areas should explain:

• what needs to improve
• what evidence shows the gap
• why it matters
• what good looks like next
• what support or follow-up will be provided

Avoid vague balancing. Do not add a development area just because the form expects one. Do not bury serious concerns inside praise. The employee deserves accuracy.

Watch for risky language

AI is particularly useful for reviewing phrasing that could create fairness or legal risk.

Be careful with language about:

• personality
• age
• health
• family responsibilities
• communication style
• confidence
• culture fit
• emotional tone
• availability
• protected characteristics

The problem is not that managers cannot discuss behaviour. They can and should. The issue is whether the review describes job-relevant evidence rather than personal judgement.

Instead of:

She can be emotional in difficult conversations.

Write:

In two customer escalation meetings, the discussion moved away from the agreed facts and next steps. Next quarter, I need you to keep escalation conversations anchored to the issue, owner, and resolution timeline.

That language is clearer and more defensible.

Add an evidence table before finalising

Before a performance review is shared, create a simple evidence table.

Use columns:

• claim
• supporting example
• impact
• confidence level
• follow-up needed

This table is not necessarily shown to the employee. It is a manager preparation tool.

It prevents the manager from presenting strong claims with weak evidence. It also helps HR spot where feedback may be based on recency bias, hearsay, or a single incident.

If a claim has no example, either find evidence or remove it.

Make the review conversation two-way

The document prepares the conversation. It is not the conversation.

Managers should invite the employee to respond, clarify, and add context. This does not mean the manager avoids making an assessment. It means the employee is treated as a participant in their development.

Useful questions include:

1. Which part of this review feels most accurate to you?
2. Where do you see the evidence differently?
3. What achievement from this period should we make sure is captured?
4. What got in the way of stronger performance?
5. What support do you need from me in the next review period?

These questions create a better conversation and may reveal context the manager should know.

Looking ahead matters as much as looking back

A performance review that only summarises the past is incomplete. Employees need a view of the next period.

The "Looking Ahead" section should define:

• priority outcomes
• behaviours to continue
• behaviours to change
• development focus
• manager support
• follow-up rhythm
• what good performance will look like

This is especially important when feedback is difficult. The employee should not leave with only a diagnosis. They should leave with a path.

Today's practice

Choose one real review. Paste rough notes into the prompt. When the output returns, do not edit for elegance first. Edit for truth.

Ask:

1. Is every claim supported by evidence?
2. Has AI softened feedback that needs to be clear?
3. Has it exaggerated praise beyond the examples?
4. Would the employee know what to repeat or change?
5. What follow-up conversation does this review require?

By the end, you should have a review draft that is more specific, more useful, and easier to discuss with confidence.