78% of managers used AI on reviews and 16% of staff were told. Write the context the tool cannot see

Highwire, the professional development programme run by Knopman Marks, published a survey on 6 October of 1,034 US full-time corporate professionals at organisations with 100 or more staff. Of the 321 people managers in the sample, 78% said they had used an AI tool to handle employee feedback or a performance review in the past 12 months. Of the 713 individual contributors, 16% said they were told AI played a role in their most recent review. Another 40% either suspect it was used without confirmation or are not sure. HR Dive and HR Executive both reported the findings in October. For anyone on a calendar-year cycle, the next review is not far off.

What the transparency headline misses

The obvious reading is a disclosure story: managers are using a tool and not saying so. The more useful reading is about inputs.

A model drafting a review works from whatever it is given. Liza Streiff, Highwire's chief executive, told HR Dive that a good manager can tell a smart risk that failed from a careless mistake, but "context is exactly what AI doesn't have."

The employee results show what that means in practice. In Highwire's data, 54% said feedback had become more specific and actionable since AI entered the process. 34% said it had become more generic, and 32% said it was less useful. Same process, opposite outcomes. Streiff put the difference down to how involved the manager stays.

Read that as a statement about raw material. A manager who pastes in a ticket count and one half-remembered incident gets a generic paragraph. A manager who has your year in front of them gets something specific.

You do not control whether your manager uses AI. You control what is on the desk when they do.

The cost of playing safe

The second set of findings matters more for careers.

Highwire found 70% of employees and 75% of managers believe making mistakes is critical to learning new skills at work. Only 46% of employees feel they can make one without it being held against them. Two-thirds of employees say that concern changed their behaviour: 27% avoided a reasonable risk, 27% delayed asking for help, 21% held back an idea and 16% hid or downplayed a mistake.

The fear has a basis. 55% of employees and 54% of managers agree that an unsuccessful risk could hurt a performance evaluation. 42% of managers say mistakes can lead them to limit future assignments for the person who made them. And 15% of workers say AI gets more room than employees to correct a poor result and try again.

Put those numbers next to a drafting tool with no context. A failed experiment recorded only as a missed target reads, in summary, much like negligence. The rational response is to stay small. That response moves you away from the work that holds its value: decisions under ambiguity, and owning the call when one goes wrong.

AI does the routine work. You do the thinking. A review process that cannot see the thinking pays people to retreat into the routine, which is exactly the work AI absorbs first.

Your Next Move

Send your own record before the review is drafted. One page, two weeks ahead of the meeting, framed as input for your manager. Three outcomes with numbers attached. One risk you took, with the reasoning at the time and what it taught you. The work nobody saw, such as the handover you rescued or the colleague you trained. If AI is drafting the review, this is the context it lacks. If it is not, you have still done half of your manager's preparation, and the record exists either way.

Log the reasoning on risks when you take them, not afterwards. A dated note with the decision, the options you weighed and why you chose one. Streiff's distinction between a smart risk and a careless mistake is made from exactly this kind of record. Written in advance, it turns a failed bet into evidence of judgement. Reconstructed at review time, it reads as an excuse.

If you manage people, say how you used AI and spend the saved time in the room. Streiff's advice to HR was to make sure managers are open about how AI was used and that the time it saves goes back into the conversation. One line does it: what the tool structured, and which assessments are yours. Then own every sentence. Your team will hold you to them.