AI performance reviews are already here. Most employees just do not know it. A new survey reported this week found that more than three in four managers have used AI to help with employee feedback or a performance review in the past year, while only a small fraction of employees were told AI played a role in their own.

We have spent two years asking employees to be transparent about how they use AI. It turns out the same question now applies to the people writing their reviews.

This is not a story about lazy managers. It is a story about trust, and about a rule that most organizations never wrote down.

Key takeaways

  • In a Highwire survey covered by HR Dive, 78% of managers said they used AI to handle feedback or a performance review in the past 12 months, but only 16% of individual contributors said they were told AI played a role in their latest review.
  • The issue is not that managers use AI. The issue is that the path behind the most consequential conversation in someone’s job has become invisible.
  • A performance review is high-consequence work, so it deserves high visibility: disclose AI’s role and keep the judgment clearly human.
  • Leaders should define AI’s role in evaluation before managers and employees define it informally.

What’s happening

On October 8, HR Dive reported on a new Highwire survey about AI in performance reviews. Highwire, a performance coaching program from the exam-training firm Knopman Marks, surveyed more than 300 U.S. people managers and more than 700 individual contributors. The Who’s Managing Whom report was released October 6.

The headline numbers are striking. 78% of managers said they had used AI tools in the past 12 months to handle employee feedback or a performance review. Only 16% of individual contributors said they were told AI played a role in their most recent review.

Employees are split on the result. 54% found AI-generated feedback more specific and actionable. About a third found it more generic (34%) or less useful (32%). And more than half of employees said they feel mistakes can be held against them. Two-thirds said they change their behavior to protect themselves, from avoiding reasonable risks (27%) and delaying requests for help (27%) to holding back ideas (21%) and hiding or downplaying mistakes (16%).

Liza Streiff, Highwire’s co-founder and CEO, put it plainly: “AI can help you write a faster review, but it doesn’t replace human judgment, context, nuance and empathy.” On telling a smart risk apart from a careless mistake, she added that “context is exactly what AI doesn’t have.”

The same day, HRD America asked a related question: is AI giving employees a distaste for effort? The experts disagreed in useful ways. Babson College’s Tom Davenport said people adopting AI to go faster are “less interested in exercising their brains.” Steven Shaw of King’s Business School, co-author of the 2026 paper on “cognitive surrender,” pointed to reinforcement: when people are rewarded for outsourcing their reasoning, the habit becomes routine. Boston Consulting Group’s Gabriella Kellerman said, “I haven’t seen any evidence of an overall distaste for effort,” though she does see over-reliance on AI for specific tasks.

Two details in that piece matter for this conversation. It cites an Omni Calculator survey finding that “54% of managers run 70% or more of what they write past AI” before sending it. And it cites Gallup data from May 2026 showing that just 25% of U.S. employees said their organization had communicated a clear plan for integrating AI.

Put those together and the picture is clear. AI is already inside the manager’s work. The rules for it mostly are not.

Why AI performance reviews are a trust issue, not a tool issue

In The Next Turn, I describe two changes happening at once. The first is the definition of work: what people believe work is for and what it should give back. The second is the structure of work: how the work actually gets done, which AI is now reshaping from the inside. A performance review sits right at the collision point.

In Chapter 9, Invisible Work, I write about what happens when AI removes the trail that work used to leave behind.

Leaders increasingly see the product without seeing the path.

The Next Turn

For two years, that sentence has mostly described leaders looking at employee work. This survey flips it. Now employees are receiving a polished product — their own review — without seeing the path behind it. Was this my manager’s assessment? A summary of my manager’s notes? A draft my manager barely edited? Most employees in this survey did not know.

That is the same uncertainty leaders feel when an employee turns in fast, polished work. And it produces the same reaction. As I put it in Chapter 9:

It is not just a workflow issue. It is a trust issue.

The Next Turn

Notice what the Highwire data says employees do when trust is thin. They avoid reasonable risks. They delay asking for help. They hold back ideas. Those are exactly the behaviors that judgment depends on. In Chapter 8, The Rise of Judgment, I argue that AI changes where human value sits, and that value increasingly sits in framing problems, reading context and knowing when the tool is pulling the work off course. A review process that people experience as opaque teaches them to hide the very thinking we most need to see.

Standards vs. signals in the review itself

Chapter 13 separates standards from signals. Standards are what the organization actually needs: sound judgment, accountability, care, follow-through. Signals are the visible cues we use to infer those things.

A performance review has always relied on a few signals of its own. The time a manager clearly spent on it. The specific examples. The sense that someone who knows your work wrote it. AI can now generate the appearance of all three. That does not make AI-assisted reviews bad. More than half of employees in this survey found the feedback more specific. It means the signal no longer proves the standard.

The standard for a review is not “my manager typed every word.” It is “my manager made an informed, fair judgment about my work and owns it.” AI can support that standard. It cannot meet it on the manager’s behalf.

This is also where the HRD America conversation matters. Shaw’s point about reinforcement applies to managers as much as anyone. If a manager is rewarded for getting reviews done faster, and nobody asks how the judgment was formed, it is easy for drafting help to become decision help without anyone choosing that. Kellerman’s caution cuts the other way, too: this is not evidence that people stopped caring. It is evidence that the expectations were never defined.

That is why I keep coming back to Chapter 12, Defining Work Again. One of the seven moves is to define AI’s role in the work before employees define it informally. When only 25% of employees say they have heard a clear AI plan, informal definitions are filling the gap. In performance management, the informal definition is apparently “use it and don’t mention it.”

Chapter 13 offers the principle I would apply here:

The amount of visibility should match the level of consequence.

The Next Turn

Few things at work carry more consequence for an individual than how their performance is judged. That is the place for more visibility, not less.

Your next turn

If you lead managers, or you are a manager writing reviews this season, here are five moves for this week:

  • Write the rule down. Decide where AI may help in reviews (organizing notes, tightening language, checking for vague phrasing) and where it may not (forming the rating, deciding promotion or pay outcomes).
  • Disclose by default. Ask managers to say, in one sentence, how AI was used in a review. Not as a confession. As a norm.
  • Keep the judgment visible. Before any review is delivered, a manager should be able to talk through how they reached their conclusions. That is the question I return to again and again in Chapter 8.
  • Check the context gap. Ask managers to name one thing about the employee’s year that no tool could have known: a smart risk that did not pay off, a quiet contribution, a hard tradeoff.
  • Make mistakes discussable. If people are hiding mistakes to protect their evaluations, the review process is teaching the wrong lesson. Say explicitly how a reasonable, well-reasoned risk will be treated.

Frequently asked questions

Should managers use AI to write performance reviews?

AI can help managers organize notes and sharpen feedback, and many employees in the Highwire survey found AI-assisted feedback more specific. The rating and the judgment behind it should stay with the manager. Treat AI as drafting support, not as the evaluator.

Do employees need to be told when AI was used in their review?

There is no single rule that covers every organization, so leaders need to set one. Given that only 16% of employees in this survey said they were told, and a performance review is high-consequence, I recommend disclosure as the default. It protects trust more than it costs time.

How do we keep AI-assisted reviews fair?

Start by defining the standards you are evaluating against, separate from the signals that used to stand in for them. Require managers to explain their reasoning in their own words and to add context only they could know. Then review a sample of evaluations to see whether the judgment is visible.

The organizations that handle this well will not be the ones that ban AI from reviews or the ones that quietly let it run. They will be the ones that say, out loud, what the review is for and who owns it.

We asked employees to show their work with AI. It is fair for them to ask the same of us.

So here is the question: if someone on your team asked how their last review was written, could you answer it plainly?

Sources