Survey: 78% of Managers Use AI for Performance Reviews, but Only 16% of Workers Are Told

Only 16 percent of non-managerial employees surveyed were told that AI had influenced their review, leaving a wide gap between how often the technology is used and how often its use is disclosed. ZDNet noted that an Atlassian study published in June found that admitting AI use at work can backfire, a possible reason for the silence.

How the tools are used matters. Having AI format or summarize observations written by a human is a different level of intervention than feeding performance data to an AI system and asking it to produce a review from top to bottom.

Employees' assessments of the resulting feedback were mixed. Fifty-four percent said feedback has become more specific and actionable. Thirty-four percent found it more generic, and 32 percent said it was less useful. A majority reported a benefit, but about a third drawing less value from their reviews than before AI entered the process is not a small share.

The survey also found that nearly one in four employees now rehearses difficult workplace conversations with an AI tool. Highwire describes the practice as filling a vacuum: only 44 percent of employees said they had their manager's support to practice a high-stakes conversation or another difficult workplace scenario, far below the 85 percent who said such practice is key.

A similar need has drawn product investment. Synthesia, which makes AI video tools, released a feature called Sessions that lets employees practice conversations with the company's AI avatars and receive an assessment from an "AI coach," an automated, scaled version of the one-on-one training managers typically provide. Synthesia told ZDNet it developed Sessions after enterprise customers asked for broader access to those managerial touchpoints without stretching human bandwidth. The company says Sessions is not a replacement for mentorship, and data on how effective the training experiences are for employees is not yet available.

Highwire's findings come with a commercial caveat. Its business depends on companies buying human-led professional training that develops employees' people skills, a mission that contrasts with Synthesia's. Its website says the company "conditions the human skills that can't be automated," and it does not treat rehearsing with AI as a valid alternative to preparation between employees and managers.

ZDNet noted that some employees might place unhelpful AI-generated performance reviews under the broader "workslop" label for low-effort AI output.