Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

3 places performance review bias hides

Large organizations must look beyond where it's expected.

[Photo: Getty Images]

Profile photo of Bora Unlu

Cofounder and CEO of Teamflect · Teamflect

If we take a look at the discourse around performance reviews today, we can see that the human resources (HR) community is more aware of all the different types of bias that can arise in performance reviews and has more sophisticated tools for bias detection at its disposal than ever before.

That said, as performance review methods evolve, the bias that remains becomes more difficult to identify. This is especially true in larger organizations. In fact, a 2025 WTW survey found that only 39% of organizations consider their performance process effective at delivering clear goals, fair ratings, regular feedback, and strong ties to rewards.

When one HR leader can read every review, unfair ratings tend to stand out. When thousands of reviews come in from hundreds of managers each cycle, bias becomes difficult to spot in any single document and easy to overlook in the aggregate.

At Teamflect, the performance management software company I co-founded, we work with many organizations at that scale, and I’ve come to see review bias in large companies as largely a process-design challenge.

So I want to focus on where we’ve found bias can hide in enterprise performance reviews and how it can be mitigated.

1. Bias builds in review history

A 2025 field study at a U.S.-based multinational financial services firm found that when a software glitch kept managers from seeing employees’ self-evaluations, they anchored on the previous year’s ratings instead. Existing gender and race gaps in ratings barely moved. The pattern changed only for newcomers, who had no self-rating history to anchor on.

One approach I’ve seen work is auditing performance review data in bulk before ratings are finalized, and regularly auditing past performance review scores for emerging trends and patterns.

2. Calibration can reward the most persuasive manager

Calibration meetings exist to correct inconsistent ratings across managers. They’ve become close to universal and a staple of performance management in enterprises. In fact, The Talent Strategy Group’s 2026 Performance Management Report found that 84.7% of organizations hold them.

While I strongly support the calibration process when applied correctly, it can also reinforce and conceal review bias.

A 2025 study of one Brazilian company found that managers were strategic about the information they shared during calibration. Supervisors who found it easier to avoid scrutiny and defend their ratings received fewer adjustments, and their employees ended up with higher ratings, even after accounting for actual performance. In practice, an employee’s final rating can depend partly on how comfortable their manager is making a case in a meeting.

In my experience, what has helped is moving the first pass of calibration out of the meeting room. Review software can now analyze every completed review form in bulk, compare each rating with the written evidence behind it, and flag where the two don’t line up. The calibration meeting then starts from those flagged cases, and the discussion stays anchored to what managers actually wrote.

3. AI changes who writes the review

Generative AI has quickly become part of how managers write performance reviews. One recent survey of U.S. professionals suggests that 78% of managers said they had used an AI tool to prepare feedback or a review in the past year, yet only 16% of individual contributors said they were told AI played a role in their most recent review.

AI can speed up feedback and, in some cases, make it more specific. My concern is accountability. Starting with a polished AI draft makes it harder to tell whether the final text reflects their own assessment or the tool’s defaults. At enterprise scale, these tools may also push reviews toward similar language, making real differences between employees, and any bias behind them, harder to detect.

One approach I’d suggest is setting clear rules before the next review cycle on which tools managers can use and when employees should be told AI was involved. Managers can use AI to organize their notes or check their language while keeping judgment and ratings their own.

Look for bias in processes

Across these three areas, the common thread is that bias in large organizations tends to accumulate inside the process itself, where no single manager can see it. Training managers to recognize bias still matters, but in my experience, it works best alongside a process built to catch what individuals miss.

For leaders preparing for their next review cycle, a useful starting point is to find where ratings depend on judgment nobody checks, then add evidence and clear rules. Fair reviews at scale come from systems that make bias visible early enough to correct it.

Bora Unlu is the cofounder and CEO of Teamflect.