Your manager is not writing most of the words in your next review. Software is drafting it first.
Almost three quarters of HR leaders now believe that's already happening on their own teams.
The Shift Nobody Voted On
Lattice surveyed more than 1,300 HR professionals across the UK, Europe and North America. The survey feeds its 2027 State of People Strategy Report, published in 2026 under Lattice's forward-dated naming convention. Seventy six percent said AI shortens how long a review cycle takes.
Sixty four percent said it improves the quality of the review itself. That's a vendor's own customer base, though. Treat the quality number as a user's impression, not an independent audit.
"Managers must remain accountable for their feedback and decisions." Sophie Hurcombe, Lattice
That line is getting harder to hold in practice. Sixty percent of the same HR leaders still have ethics concerns. The concern is how the technology actually gets deployed.
Why Companies Are Racing Ahead
Three forces are pushing this, not one. Speed is the easy sell.
Lattice's own product team says early customers finished reviews 47% faster from a generated draft. That's a vendor claim, not an independent benchmark.
Consistency is the second force. A draft built from the same data fields reads the same way across many employees. It doesn't vary with a manager's writing habits, or their mood that week.
Legal defensibility is the third force, and companies talk about it least.
The Hackett Group's 2026 HR Key Issues Study has a number for this. Talent and performance management AI adoption lands at 42%. That's by year end.
Eighty one percent of HR leaders in that study see real value in the technology. Yet 59% still have no rollout plan.
The Legal Logic Gets Complicated
This is an AI governance gap, not just a drafting problem. Compliance analysts Hekim Colpan and Phillip Wikes gave the risk a name: decision reconstruction risk.
"A consequential record can no longer show, on its own, why a decision was made." Hekim Colpan and Phillip Wikes, Corporate Compliance Insights
Their warning is specific. An AI-drafted review can look more polished than the evidence underneath it. The file reads cleaner than the actual record supports.
Tucker Law reviewed a Resume Builder survey of more than 1,300 US managers. All of them have direct reports.
About six in ten already rely on AI for decisions about those employees. That includes raises for 78% of them, and performance assessments for 91%.
More than one in five managers said AI frequently makes the final call. No human checks it first. Two-thirds using these tools have had no formal training on them at all.
"The 'black box' nature of many AI systems makes it, at best, difficult for an employer to fully explain." Tucker Law, on AI-influenced performance decisions
That opacity is exactly what a discrimination claim asks an employer to resolve.
What Employees Already Suspect
Kickresume surveyed 1,365 employees across six regions. Seventy two percent suspect their manager used AI to write their review.
Among those who suspect it, 30% said parts felt generic or overly polished. Eighteen percent believed the whole document was mostly or entirely AI-written.
Gartner's numbers go further, and they are not about suspicion. They surveyed nearly 3,500 employees in October 2024.
Eighty seven percent said an algorithm would probably be fairer. Fairer than their manager.
"A damning indictment of how employees feel about managers." Emily Rose McRae, Gartner
(no AI draft)
(manager skims, approves)
(rubber-stamped)
(reviewed and defended)
Positioning is Aevum Transform's analysis of the sourced findings above, not a measured dataset. Sources: Lattice product updates, Tucker Law.
What a Draft Can't See
JPMorgan Chase now lets managers use an internal chatbot to help draft reviews. The bank's guidance says the tool should not set scores or decide pay and promotion.
Wharton management professor Peter Cappelli has a point about this. Human managers already overweight recent events.
They also bring their own bias into reviews. On that measure, a model is not automatically worse.
But these tools tend to agree with whoever is using them. That draft misses something important.
It won't flag what a manager catches face to face. Someone quietly disengaging.
Someone covering for a struggling teammate. Someone about to walk.
That's the context a dataset doesn't log. Nobody wrote it down anywhere the system could read it.
The Coaching Counterweight
None of this argues for banning the drafting tools. The 47% speed claim and the consistency argument are both real reasons to want one.
It argues for training the judgment that has to sit on top of the draft. Reading a room is one example. Asking the follow-up question is another.
So is defending a rating out loud when legal sits in the meeting. None of that comes bundled with the software.
That's the gap executive coaching exists to close. It's also where East Valley companies are putting their 2026 leadership development budget. Not into another HR platform license.
The draft can write the sentence. Only a trained manager can defend it in the room.
Start a Conversation →Score Your Review Process
Run your own process against the seven questions below. Each "yes" is a point toward judgment. Judgment your company could actually defend if a review got challenged.
How to use this: Check every statement that's true of your company's current review cycle. Then see your score.
Nothing here is saved or sent anywhere. It only runs in your browser.
- A manager reads and edits every AI-drafted review before it's finalized.
- The draft cites specific examples, like named projects or dates, not generic language.
- Employees are told when AI assisted in writing their review.
- Someone audits AI-drafted ratings for patterns across teams or demographics.
- The manager can explain, in their own words, why each rating was given.
- Difficult conversations, like a PIP or termination, still happen face to face.
- Managers received real training on the AI tool, not just a login.
This is a directional self-check, not a legal compliance audit. Consult employment counsel for an actual risk assessment.
Watch whether your next review cycle still produces a sentence your manager could defend without the software open in front of them. If it can't, you have a drafting tool making personnel decisions, not a manager using one.
Frequently Asked Questions
How many companies are using AI to write performance reviews in 2026?
Almost three quarters of HR leaders believe managers already use AI. They use it at their own company, to draft reviews.
That's per Lattice's 2027 State of People Strategy Report. The Hackett Group separately projects 42% HR adoption in talent and performance management. That's by year end.
Do employees know when their performance review was written with AI?
Most suspect it. A Kickresume survey of 1,365 employees found 72% suspect their manager used AI. AI to generate their review.
Of those, 30% said parts felt generic or overly polished. Eighteen percent believed the review was mostly or entirely AI-written.
Is it legal to use AI to write employee performance reviews?
Drafting with AI is generally legal. Compliance analysts warn it creates decision reconstruction risk instead. A polished record that can't show why a decision was actually made.
That becomes a liability if a discrimination claim asks an employer to explain its reasoning.
Do employees trust AI more than their manager for performance feedback?
Many say they would. A Gartner survey of nearly 3,500 employees found 87% believed an algorithm would be fairer. Fairer than their own manager.
Gartner's own analyst called that finding a damning indictment of how employees feel about managers.
A faster review isn't the same as a defensible one.
Coachvox AI helps you build the judgment an AI draft can't supply. Not another memo to explain.
Affiliate disclosure: This page contains affiliate links to Coachvox AI. If you purchase through these links, we may earn a commission at no additional cost to you. See our full disclosure policy.
Explore Coachvox AI →