ZDNET’s key takeaways
- Managers are increasingly using AI to provide feedback to direct reports.
- How employees receive this is a mixed bag.
- AI fills a support gap in conversation prep.
How AI uses ramps at work, where are the advantages and disadvantages, and for whom? New research from professional development firm Highwire shows where AI has entered the feedback loop—and where it’s both a potential problem and a solution for employee development.
Highwire surveyed 1,034 company officials on how AI is infiltrating manager-employee relationships. It found that 78% of managers reported using AI to “help design, edit, summarize or inform feedback or a performance review” in the past year. Most of these managers, perhaps not surprisingly, noted that AI has accelerated the process of preparing this feedback (Highwire told ZDNET that it did not ask managers what else they noticed about the use of AI).
Also: Almost 70% of workers now regularly use AI – without time to qualify
However, only 16% of non-manner employees who were surveyed were told that AI affects their assessment. Perhaps concerns about being perceived poorly for using AI are behind that drop in transparency: an Atlassian study from June found that AI use at work has backfired.
The details of how managers use AI for feedback are important here, however. Having an AI tool format or summarize human-written observations, for example, is a different level of intervention than handing AI some performance data and asking it to issue a top-to-bottom review.
How that feedback lands with employees: 54% of respondents said feedback “has become more specific and actionable.” But 34% found the feedback more generic, and 32% said it was less useful. A third of employees potentially deriving less value from their assessments than before AI came into the picture is not small, even if the majority finds an advantage.
Also: What workers will really be using AI for in 2026—and what they won’t be
The survey also found that “nearly 1 in 4 employees now rehearse difficult workplace conversations with an AI tool.” Highwire frames this as the result of a vacuum: Only 44% of employees reported that they have their manager’s support to practice a high-stakes conversation or other difficult workplace scenario, far less than the 85% who say that practice is key. AI therefore fills a need space for nervous employees.
AI tools to the rescue
The results of the trials with AI are in the AI video company Synthesia’s latest release, Sessions. The new feature allows employees to practice conversations with one of the company’s AI avatars and receive an assessment from an “AI coach” — essentially an automated, scaled version of the one-on-one training managers usually give employees.
Synthesia told ZDNET it developed the feature based on feedback from enterprise customers who want to expand access to these managerial touchpoints without straining human bandwidth (although the company says Sessions is not a substitute for mentorship). For Synthesia customers, a tool like Sessions fills in training gaps they were previously unable to fix (although data on how effective these training experiences are for employees remains to be seen). Both Highwire’s take and Synthesia’s investment in sessions point to a simulation trend at work, with AI playing a bigger role as employees learn.
Also: Let AI ‘do things no one wants to do’ – and 4 other ways to use agents more effectively at work
It’s worth taking Highwire’s survey intentions with a grain of salt. Much the opposite of Synthesia’s mission, Highwire’s business relies on companies that need human-led professional training that invest in their employees’ skills. That does not necessarily involve teaching managers how to expand responsibility for AI, and it does not see trials with AI as a valid alternative to employee-manager preparation.
“Highwire conditions human capabilities that cannot be automated,” reads the company’s site.
Still, the takeaway isn’t far off from other, less-than-desirable AI job trends. Some employees may argue that unhelpful AI-generated performance reviews fall under the Worklop umbrella.