Should we put AI in our performance reviews?
Question
Should we put AI in our performance reviews?
Every HR-tech vendor in 2026 is pitching an AI feature on top of performance management — review drafting, peer-feedback summarization, capability scoring, manager coaching, AI-usage tracking. Adopting something is the easy decision; where to place it in the review process is the hard one.
The 2026 evidence is unusually clear because two large platforms went in opposite directions in the same quarter. Meta added employee AI usage as a performance signal; Duolingo dropped it after public backlash. Both companies have data. They reached opposite conclusions. This page is for HR leaders deciding which side to land on — and which uses of AI are safe regardless.
Evidence
These findings frame the decision; the last two add legal and implementation depth.
Duolingo reversed course publicly. Entrepreneur covered the about-face: Duolingo's CEO mandated AI use as a performance metric, sparked sustained employee backlash, and changed the policy within weeks. Storyboard18 reported the implementation detail — AI usage is no longer a performance metric at Duolingo. The signal: even at an AI-native company, tying AI usage to evaluation outcomes was net-negative on internal trust.
HR Executive's expert verdict: "spectacularly bad." A direct critique from HR Executive labels AI-usage incentives — bonuses, ratings, promotion criteria — a "spectacularly bad idea." The reasoning: tying compensation to a tool's usage rewards theater over outcome, distorts adoption patterns, and breaks the link between performance and impact.
Meta is running the opposite experiment. Fortune reported in January 2026 that Meta now links employee AI usage to performance reviews and rewards via an "impact-evidence" framework. The Checkpoint system enforces a forced-distribution model (roughly 20% Outstanding / 70% Excellent / 7% Needs Improvement / 3% Not Meeting Expectations), with bonus multipliers up to 300% for top performers via the new "Meta Award" tier. For software engineers specifically, the model tracks AI-generated code lines among 200+ data points. The honest read of this disagreement: two large platforms with real data reached opposite conclusions in the same six-month window. We don't yet have outcome evidence to call a winner.
Cornell research: narrative reviews are perceived as fairest. A Cornell Chronicle study found that narrative-format reviews — written, evidence-grounded, story-shaped — are rated fairest by employees. Implication: AI is genuinely good at first-draft narrative writing and synthesis, which maps onto the review format employees trust most. The question isn't whether AI can help. It's whether it can help without becoming the judge.
People Management cataloged the bias surface. People Management's "what HR needs to know" brief flags the audit, transparency, and consent issues: AI trained on past ratings inherits past bias; reviewer trust drops sharply when employees can't tell whether a human or model wrote what they're reading. JD Supra's legal-risk guidance reinforces this — AI-assisted performance decisions create new disclosure and discrimination-claim surface area that most HR policies don't yet cover.
Disagreement
The corpus disagrees in three predictable ways.
Vendors and consultants read the technology shift as inevitable and pro-productivity — the bulk of the 2026 AI-in-PM market-growth coverage falls here. Practitioner press (HR Executive, People Management) is cautious, focused on second-order effects on trust and bias. The companies themselves are running opposing experiments — Meta and Duolingo same quarter, opposite verdicts. Academic research (Cornell) supports AI assistance in narrative drafting but says nothing about AI judgment.
So the live disagreement isn't "AI yes or no." It's three sequential questions, and the evidence pattern is different for each:
- Should AI generate review content (drafts, summaries) — where it operates as a writing aid? (Evidence converges: yes, with disclosure.)
- Should AI score employees against criteria — where it operates as evaluator? (Evidence is mixed-to-negative.)
- Should AI usage itself be a performance signal — where the tool becomes the outcome? (Evidence is sharply negative.)
Most vendor pitches blur questions 1, 2, and 3 into a single product story. The HR leader's job is to keep them separate.
Peoplense Verdict
Do: use AI as a drafting and synthesis aid — first-pass review writing, peer-feedback summarization, evidence aggregation for the manager. Cornell's research suggests well-structured narrative output may actually raise perceived fairness. Keep the human as the editor and decision-maker. Disclose AI involvement as a matter of policy, not on request.
Don't: use AI as the scorer without rigorous bias validation on your own data. AI trained on past ratings inherits past bias — exactly as the Harvard Kennedy field study showed for human reviewers — except now the bias is opaque, scalable, and harder to challenge. And don't tie AI usage to ratings or compensation. HR Executive's expert verdict on this is unambiguous, and Duolingo's reversal is the live proof.
Watch out: AI-assisted reviews that aren't disclosed as AI-assisted destroy trust faster than slow reviews. The fairness perception is asymmetric — employees rate human reviews higher when they know they're human, and rate AI reviews far lower when they discover they're AI without being told. The disclosure rule matters more than the technology choice.
What to do Monday
Three concrete actions for HR leaders this week.
-
Audit your current AI-in-reviews exposure. List every tool in your review process that uses AI — Workday/SAP-built features, Lattice/15Five summaries, embedded ChatGPT, vendor-trained capability scoring. For each, identify which step it touches (draft / summary / score / recommendation). Most teams have more AI in the review process than they realize, often via vendor features quietly enabled in the last update.
-
Write a one-page AI-in-Reviews disclosure policy. Three lines: what AI does, what the human does, how employees can see and contest the AI portion. Communicate it before the next review cycle opens. The Cornell + People Management evidence converges on the same point: transparency is the trust differentiator, not capability.
-
Kill any AI-usage performance metric you have or are considering. This is the highest-leverage move on the page. HR Executive's verdict is unambiguous and Duolingo's reversal is the empirical lesson. Measure outcomes — project impact, customer outcomes, peer trust — not tool usage. If you genuinely need to incentivize adoption, do it with training time and tool access, not with ratings or compensation.
Data Privacy — The Legal Dimension
Any conversation about AI in performance reviews eventually hits one question that can't be sidestepped: what does the law say about employee data?
This isn't an ethical concern alone. It's a legal obligation with clear penalties in most jurisdictions.
Globally, major data-protection frameworks — GDPR in the EU being the most consequential — set explicit rights for employees regarding automated decision-making. The most important among them:
- The right not to be subject to a decision based solely on automated processing if it produces legal or similarly significant effects, without meaningful human review.
- The right to meaningful information about the logic involved in the decision.
- The right to contest the processing and request human review.
If your organization is using an AI model to evaluate employees or drive compensation, you need:
- A clear lawful basis for processing. Consent in an employment relationship is often insufficient — employees can't freely refuse when their job depends on it.
- Up-front disclosure in the employee privacy notice, before the tool goes live.
- A documented contestation channel.
- A Data Protection Impact Assessment (DPIA) before deployment, documenting risks and mitigations.
In Saudi Arabia, the Personal Data Protection Law (PDPL) — in force since 2024 — adds parallel obligations with local-specific dimensions:
- Cross-border data transfer is tightly regulated. Most major AI vendors host data on servers outside the Kingdom, which means using a cloud AI service on Saudi employee data is a legal-review question, not just a procurement decision.
- Employee consent within an employment relationship faces the same scrutiny as under GDPR — when refusing could affect career outcomes, the consent's validity is questionable. Free, informed consent is the standard, not just a signature.
- Performance data often intersects with sensitive categories (health, absence patterns, prior ratings) — triggering higher protection requirements.
Bottom line on this dimension:
Even if the tool is technically capable, it may not be legally permissible — especially uses where AI is the final judge of performance.
Handle this before deployment, not after.
GCC Relevance
The KSA / GCC workplace adds three specific pressures to this decision.
Vision 2030 incentivizes AI investment — which makes vendor pitches harder to refuse. Government accelerators, GASTAT digitization mandates, and corporate AI-readiness scorecards pressure HR teams to deploy AI fast. The risk: rushed AI-in-PM rollouts skip the disclosure and validation steps, then surface bias issues that the labor courts haven't yet ruled on. Slower adoption with documented evidence is the defensible posture.
Power-distance amplifies the disclosure problem. Employees in high-power-distance contexts are less likely to challenge an AI-generated rating they suspect is unfair — the same dynamic that produces "Meets" inflation in human reviews. Without proactive disclosure and a contestation channel, AI in reviews silently encodes the existing trust gap rather than narrowing it.
Wasta dynamics make AI-judgment particularly dangerous. If the model is trained on past ratings — which in any GCC workplace will partially encode relational politics — the AI institutionalizes wasta in a way that's harder to detect and challenge than the human version. AI as drafting aid stays safe; AI as judge, in this context, is riskier than the human baseline it would replace.
Practical implication for KSA HR leaders: the question for most Gulf workplaces isn't "should we adopt AI in reviews" — vendor and government pressure makes some adoption near-inevitable. It's "can we keep AI as drafting aid, not as judge, while everyone around us is selling us the judge?" Answer that first.
Sources
All corpus articles below open in our admin reader with the editorial-summary contract banner — our text summary on Peoplense, full text and figures at the original publisher.
- Duolingo's CEO Sparked Backlash Over Performance Reviews — Now He's Changing Them — Entrepreneur, 2026-04-14 (the about-face story)
- Duolingo drops AI usage metric in employee performance reviews — Storyboard18, 2026-04-15 (implementation detail)
- Incentives for AI use: A "spectacularly bad idea" — HR Executive, 2026-04-17 (direct expert critique)
- Meta is changing its performance review to reward output over effort — Fortune, 2026-01-13 (the opposite-direction case)
- Narrative-based performance reviews deemed fairest by employees — Cornell Chronicle, 2026-01-05 (research on perceived fairness)
- AI in performance reviews: what HR needs to know — People Management, 2026-03-27 (practitioner overview of bias / consent / transparency)
- How Employers Can Manage Risk When Using AI for Employee Performance Management — JD Supra, 2026-03-04 (legal-risk lens)
Get the Monday Brief
Evidence-based people development research, summarized weekly. Free. No ads. Every article links to its source.
Email used only to deliver the brief. Unsubscribe anytime.
Want to join the editorial team?
We're building this slowly, with HR practitioners across the GCC. If you'd like to help shape what gets published, tell us about yourself.
Tell us about yourself