The best AI performance review examples pair a specific rating level with concrete evidence: a metric, a dated project, a named behavior. AI can draft that language in seconds, but the draft is only as good as the facts you feed it. Below are twelve sample reviews written for the situations founders and COOs face, followed by a prompt method, a comparison of twelve review tools, and twelve Hatchproof options that read the work itself.
1. How to ask AI to write a performance review
To ask AI to write a performance review, give it the role, the rating level, three to five dated facts, and the tone you want, then ask for a draft you will edit. A vague prompt returns vague praise. A prompt with evidence returns something you can defend in a conversation.
Use this structure:
- Role and level: "Senior account executive, second year."
- Evidence: "Closed $410K against a $350K quota. Missed two renewals because of late follow-up."
- Constraint: "No adjectives without a supporting fact. Under 150 words."
- Output format: "Strengths, one growth area, one goal for next quarter."
Always replace anything the model invents. If a draft cites a project you never mentioned, delete the sentence. The Hatchproof guide to AI performance reviews covers where AI drafting helps and where it introduces risk.
2. Outstanding: sample review for a high-output contributor
An outstanding review names the measurable result, the behavior behind it, and the effect on the team. Here is a sample:
"Maya rebuilt the onboarding pipeline in Q2, cutting time-to-first-deploy for new engineers from nine days to four. She documented the process, trained two teammates to maintain it, and raised the three blockers that would have delayed the Q3 release. Her decisions consistently reduced rework for others. Next cycle, I'd like her to own the cross-team tooling roadmap."
Why it works: one number, one dated project, one effect on colleagues, one forward goal. Nothing in it relies on personality.
3. Exceeds expectations: sample review for a reliable senior performer
An exceeds-expectations review credits consistent results above target while naming one specific stretch area. Sample:
"Daniel exceeded his pipeline targets in three of four quarters and mentored two new hires to full productivity ahead of schedule. His forecasting accuracy improved from 71% to 88%. To reach the next level, he should delegate more of the routine account reviews and spend that time on two strategic accounts."
The tone is warm but factual. Avoid "great attitude" unless you can name the moment it showed.
4. Meets expectations: sample review for a steady contributor
A meets-expectations review confirms the baseline was delivered and defines what moves the person up. Many managers write this one badly, so be explicit:
"Priya delivered all assigned campaigns on time and within budget this year. Her reporting is accurate and her stakeholders trust her numbers. To move toward exceeds, I'd like to see her propose one experiment per quarter and share the result with the wider team."
This level is not a criticism. It describes dependable work, and the growth step gives it direction.
5. Needs improvement: sample review that stays specific
A needs-improvement review states the gap against a stated expectation, gives two or three examples, and sets a dated plan. Sample:
"Three of the last five project handoffs reached the next team with missing documentation, which delayed delivery by an average of two days each. The expectation is a complete handoff checklist at every transition. We will review the next four handoffs together, and I expect checklist completion on all of them by the end of the quarter."
Gallup's survey of U.S. employees found that only a minority of workers strongly agree their review process is fair and transparent, per Gallup's survey of U.S. employees. Specific, evidence-based language is the most direct way a manager can move that number.
6. Unacceptable performance: sample review tied to documented expectations
An unacceptable rating needs documentation, prior conversations, and a clear consequence. Sample:
"Despite written expectations shared in February and follow-up conversations in April and June, quarterly deliverables were missed in each period and two client escalations resulted. A formal improvement plan begins on the first of next month with weekly checkpoints. If the agreed targets are not met by the end of the plan, we will discuss next steps."
Involve whoever handles employment decisions before delivering a rating like this. AI should not draft it unsupervised.
7. Remote and async teams: sample review built on written output
A remote review should point to written, observable work rather than visibility. Sample:
"Tomás's design briefs are consistently clear enough that engineering rarely needs a follow-up question. He responded to cross-time-zone requests within the agreed window in nearly every case and resolved a long-running disagreement about the component library in a single written proposal."
For distributed teams, evidence comes from documents, decisions, and handoffs. Presence in a video call is not evidence.
8. Collaboration and communication: sample review
A collaboration review cites how the person made other people more effective, with a concrete moment. Sample:
"Whenever the support team escalated a bug, Elena acknowledged it the same day and told them what to expect. In Q3 she ran a short weekly sync between support and engineering that reduced repeat escalations. Her written updates are brief and complete."
Gallup reports that meaningful weekly conversations can be short and can cover goals, collaboration, recognition and strengths, per Gallup's 2025 workplace reporting. A review that cites weekly evidence is easier to write than one built on memory of the last six weeks.
9. Review of an employee who works with AI agents
A review for someone who works alongside AI agents should judge what the agent layer returned, not how often it was used. Hypothetical example:
"Hypothetical example: an analyst who configures agents for invoice matching. Judge the agent layer by outcome, cost, and oversight: how many invoices it clears without review, whether a person audits a sample, and whether close time improves net of agent cost."
The pattern is what matters: outcome, cost, and oversight. The Hatchproof breakdown of AI agent performance measurement explains how to evaluate agent output alongside human output.
10. Manager performance review: sample for a team lead
A manager review should test results through the team: output, retention, and the quality of decisions. Sample:
"Under Jordan, the support team's backlog fell substantially while headcount stayed flat. Jordan holds regular one-to-ones. Where queue prioritization decisions were reversed, I'd like to see more input from the team before changes ship."
For more on this level, see the guide to performance evaluation for managers.
11. Goal-setting paragraph: sample forward-looking section
A goal paragraph should state the outcome, the measure, and the date. Sample:
"By the end of Q1, reduce average proposal turnaround from six days to three. Measure: median days from request to sent proposal, reviewed monthly. Support: a shared template library and a weekly 20-minute review with me."
Gallup reports that most employees formally review performance goals with their manager once a year or less, and that fewer than half strongly agree they know what is expected of them at work, in the same Gallup study. A goal paragraph this concrete closes part of that gap.
12. Short positive feedback: sample recognition note
A recognition note works when it names the action and its effect in two sentences. Sample:
"Thank you for catching the pricing error before the deck went out on Thursday. It spared us a corrected-quote conversation with the client and kept the renewal on track."
Gallup and Workhuman found that employees reporting weekly manager feedback and recognition were more often engaged, according to Gallup's research on feedback and recognition. Notes like this one are cheap to write and are best sent the week the work happens.
How AI is used in performance management
AI is used in performance management to draft review language, summarize feedback, flag patterns in work data, and prepare managers for conversations. The risk is that a model polishes weak evidence into confident prose. Three controls help: require that every claim cite a source, review drafts for bias, and keep a person responsible for the final rating. The Hatchproof piece on reducing bias in performance reviews gives practical checks.
Tools that run review cycles: a comparison
Four established platforms handle the administration of reviews. They suit organizations that want structured cycles and forms. Pricing is approximate and may have changed; check each vendor's site.
| Tool | Approx. price | Core strength | Suits |
|---|---|---|---|
| Lattice (Performance) | $10 per user/month; $4,000 annual minimum | Reviews, with goals as a separate $8 module | Reviews and goals from one vendor |
| 15Five (Perform) | $11 per user/month | Manager, peer and upward feedback | Manager-led reviews and coaching |
| PerformYard | $5 to $15 per person/month | Flexible cycles, 360 feedback, goals | Configurable review formats |
| Betterworks (Performance) | $8 per user/month; $9 with calibration | Goal-linked reviews, calibration | Larger organizations needing calibration |
These tools collect and organize opinions. None of them tells you what a person produced or what an agent returned, which is the question a founder deciding on a req needs answered. If you are weighing a switch, the Lattice versus Hatchproof comparison and the Betterworks alternatives overview lay out the differences.
Hatchproof options for evidence-based reviews
Two Hatchproof products supply the evidence that makes the samples above specific.
Hatchproof Performance Management
Performance Management gives leaders a live dashboard built from real work data rather than surveys or manager opinion. It shows who drives output, how team velocity and individual contribution move, and how talent decisions shift revenue per employee. Hatch surfaces patterns to leaders and does not rate, rank publicly or make the decision. It reads communication, execution, decisions and collaboration from the tools your team already uses, with no surveys to administer. The platform is SOC 2 Type II certified, and your data is never used to train AI models. Pricing is not listed here, so check the product page.
Best fit: founders and COOs who want review inputs drawn from work, not memory.
Hatchproof 1:1s & Team Cadence
1:1s & Team Cadence extracts takeaways, sentiment shifts and coaching insights from Zoom, Meet and Slack conversations, so one-to-ones feed the review record. It also offers AI communication coaching adapted to each person's style. Individual biometric data stays visible only to the individual, never to their manager or the organization.
Best fit: managers who run weekly conversations and want them to produce a usable record.
FAQ
What should I write in an employee performance review?
Write the results against the agreed goals, the behaviors that produced them, and one growth area with a dated next step. Support each claim with a number, a project name or a date. Keep adjectives to a minimum.
What are examples of positive review phrases?
Strong phrases pair an action with an effect: "reduced handoff errors by documenting the process," "resolved the vendor dispute in one written proposal," "mentored two new hires to full productivity." Each one gives the reader something verifiable.
Can AI write a performance review without a manager?
It can produce a draft, but a manager should supply the evidence and make the final judgment. AI cannot know what happened unless told, and it will fill gaps with plausible fiction. Treat the output as a first draft.
Your next step is to take one recent review you wrote, strike every sentence that lacks a number, date or named project, and note what evidence you would need to replace it. If you want that evidence to come from the work itself, including what agents returned against what they cost, Book a Demo with Hatchproof and see output per person across your human and agent workforce, with no surveys, no self-reports and no migration. You can start with Hatchproof Performance Management to see the dashboard that supplies it.
.png)