Should AI use be tracked in performance reviews? Experts say the metric alone 'doesn't encourage the right behaviours'
The News
Almost every Fortune 500 company is now tracking overall employee AI use, and many are going further by tying AI adoption metrics directly to performance review outcomes. But according to Kate Jensen, director analyst at Gartner, looking at AI use alone "doesn't encourage the right behaviours." The risk, she and other experts told HR Dive, is that usage targets incentivise volume over quality — what organisational psychologist Leslie Caputo at coaching platform EZRA calls AI "workslop": poor-quality outputs generated to meet a metric rather than to produce value.
Caputo is blunt about the motivation: "Companies have invested enormous amounts into AI, and they are under pressure to deliver ROI. Usage numbers are an easily achieved metric that leadership can show the board and say, 'look, we're adopting it.'" But the approach carries hidden costs. Jensen warned that tying AI use to performance evaluations can amplify employee anxiety about job replacement and "come across as a form of surveillance," as Caputo put it, eroding the very human capabilities — curiosity, discernment, willingness to challenge bad AI outputs — that create value in an AI-augmented workplace.
The alternative, Jensen suggested, is to set AI-related goals that are evaluated developmentally rather than punitively. Caputo added that "the differentiators that turned AI into real performance included discernment, curiosity, connection and humility. Smart companies will ultimately shift to measuring how employees execute strong judgement in AI usage, instead of usage or tokens alone."
Why It Matters
For TA leaders, this conversation matters because the teams you hire today will be evaluated under the performance frameworks being designed right now. If your organisation is among those tying AI usage to performance reviews, the message it sends to candidates — particularly those in knowledge-work roles where AI augmentation is expected — is part of your employer value proposition, whether you intend it to be or not. A candidate who learns that their future employer tracks AI keystrokes and ties them to bonus outcomes may draw conclusions about autonomy and trust that affect offer acceptance.
There is also a downstream talent risk: studies increasingly point to heavy AI reliance leading to skill atrophy, including deterioration of critical thinking. If your performance system rewards volume of AI use without measuring quality of judgement, you may be systematically promoting the wrong behaviours and, over time, hollowing out the very human capabilities that differentiate your workforce.
Taking Action
All action is situated in your corporate context. These are starting points, not prescriptions — choose what fits.
Connect
Who to talk with in your enterprise
Your head of people analytics or HR technology — ask what AI usage data is currently being collected, whether it feeds into any performance management dashboards, and what the stated rationale is. If usage is being tracked, request evidence that the metric correlates with business outcomes rather than simply demonstrating adoption.
Your head of learning and development — design a pilot programme that measures AI impact through output quality and peer review rather than usage volume, and compare the results against any existing usage-based metrics your organisation currently tracks.
Inspect
What to take a closer look at
Review your current performance review template or HRIS configuration: search for any fields, goals, or competencies that reference AI usage, AI adoption, or technology utilisation. If they exist, trace how the data is collected and whether it is used evaluatively (tied to compensation or ratings) or developmentally (used for coaching conversations only). Specific tool to look at: Your HRIS performance management module — check goal-setting templates, competency libraries, and any configurable fields in Workday, SuccessFactors, or your equivalent platform for AI-related metrics.
Reflect
What to think about more deeply
The most interesting question isn't "should we track AI use?" — it's "if we measure AI use and reward it without measuring AI judgement, are we building a workforce that knows how to prompt, or one that has forgotten how to think?"