Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Is Automating Your Performance Reviews
Transcript
- Lucas: So it's mid-2026, and most of us have probably had at least one performance review this year. But here's a number that might surprise you: according to a Gartner survey published last fall, 38 percent of large companies now use AI to generate at least part of an employee's written performance evaluation. Luna: Thirty-eight percent. That's more than a third of big firms. And two years ago that number was what, like twelve percent? Lucas: Exactly twelve. So we're seeing a pretty rapid adoption curve. And it's not just about auto-completing a text box — these systems are analyzing manager comments for biased language, suggesting specific examples from the employee's activity data, and even prompting managers in real-time during one-on-ones. Luna: That real-time prompting piece is interesting. So the AI is essentially coaching the manager while they're still talking to the employee? Lucas: Right. Some tools from platforms like Lattice and Culture Amp now offer a feature where during a one-on-one, the manager gets a subtle nudge on their screen — 'You haven't discussed career growth this quarter' or 'Consider acknowledging the project they led in March.' It's designed to make the review less of a once-a-year scramble and more of a continuous conversation. Luna: But doesn't that risk making the feedback feel… manufactured? Like, if the AI is telling you to say something, does it still count as authentic? Lucas: That's the tension, for sure. The vendors argue they're just helping managers remember things they already know but forget to write down. And there's data that suggests structured prompts actually reduce recency bias — you know, the tendency to evaluate someone based on their last two weeks rather than the whole year. Luna: Recency bias is huge. I've had reviews where my manager basically only talked about the project that finished the week before. So if AI can help spread the attention across the full review period, that's a win. Lucas: Yeah, and that's what a lot of companies are betting on. Workday, for instance, has a module called 'Workday Peakon Employee Voice' that feeds engagement survey data into the review draft. So if an employee said they felt undervalued in the last survey, the system flags that to the manager as a topic to address in the review. Luna: That sounds useful in theory. But what about the legal side? If an AI writes part of a review that later becomes evidence in a discrimination case, who's accountable? Lucas: Great question. And it's not hypothetical. In 2024, there was a case in the UK where an employee challenged a negative review that was partly ai generated. The tribunal ultimately held the manager responsible, but it raised questions about transparency. Some companies now include a disclaimer in the review that says 'Portions of this evaluation were drafted with AI assistance.' Luna: And do employees know that? Or is it buried in the fine print? Lucas: That varies. Some firms disclose it upfront. Others treat the AI as just another tool, like spellcheck. But I think as adoption grows, we'll see more regulatory pressure to be transparent. Luna: Speaking of transparency — and this might be a bit of a tangent, but it connects — it reminds me of why we keep this podcast ad-free. We don't have sponsors telling us what to cover or how to frame it. And that kind of independence is only possible because of listeners who support us directly. Lucas: Yeah, exactly. If today's conversation about AI in performance reviews gave you something useful, and you want to help us keep digging into these topics without any advertiser influence, you can support the show at buy me a coffee dot com slash fexingo. Luna: It's a small way to ensure we stay focused on what matters — real data, real examples, and honest discussions. No fluff. Lucas: Alright. Back to the review room. So one of the more controversial applications is using AI to analyze the language managers use in reviews for signs of bias. For example, studies have shown that women are more likely to receive feedback about 'communication style' while men get feedback about 'technical skills.' AI can flag those patterns across an entire organization. Luna: And then what? The HR team gets a report that says 'Your managers are biased'? That feels like it could backfire if not handled carefully. Lucas: Right. The tool alone doesn't fix anything. But companies like Textio have built platforms that give managers real-time suggestions — like 'This word has been shown to correlate with lower performance ratings for women. Consider using a more neutral term.' It's essentially a writing assistant for reviews. Luna: So it's like Grammarly but for equity. Lucas: Exactly. And the early results are promising. A 2025 study by researchers at Stanford and the University of Chicago looked at a large tech company that implemented an AI bias nudge during their review cycle. They found that the gender gap in performance ratings narrowed by about 15 percent after one year. Luna: Fifteen percent is meaningful. But does that come at the cost of accuracy? I mean, if the AI is nudging managers to soften language, could that inflate ratings? Lucas: That's a real risk. Some critics argue that bias-detection tools might actually encourage managers to give everyone a four out of five to avoid scrutiny. But the Stanford study didn't find that — they controlled for overall rating distribution and saw the gap close without a significant shift in average scores. Luna: Interesting. So it's not about dumbing down the feedback, but about making it more consistent. Lucas: Exactly. And consistency is the whole selling point for AI in reviews. Human managers are inconsistent — they're tired, they're busy, they have favorites. An AI system can ensure that every employee gets evaluated against the same criteria, with the same level of detail. Luna: But criteria are set by humans. So if the criteria themselves are flawed, the AI just automates those flaws at scale. Lucas: Right. Garbage in, garbage out. And that's a key point: the AI is not a magic solution. It's a mirror that reflects the values and biases of the people who set it up. So if a company values 'visibility' over 'impact,' the AI will reinforce that. Luna: What about the employee side? Are there tools that help employees prepare for reviews using AI? Lucas: Yes, and that's actually a growing segment. Platforms like BetterUp and CoachHub — which are more coaching-focused — now offer AI that analyzes an employee's own notes and calendar to suggest talking points for their self-assessment. It can pull out metrics like 'You attended 90 percent of team meetings this quarter' or 'You submitted 15 code reviews.' Luna: That could help quieter employees who don't naturally advocate for themselves. But I also wonder if it leads to a kind of metric-obsession, where employees only focus on things the AI can quantify. Lucas: That's a real concern. If the AI only tracks what's in your calendar or your ticketing system, then intangible contributions — mentoring a junior colleague, improving team morale — get left out. Some companies are tackling this by allowing managers and peers to submit 'kudos' or 'shout-outs' that feed into the AI's data set. Luna: So it becomes a richer picture. But also potentially a creepier one, if the AI is aggregating data from every Slack message and calendar invite. Lucas: Yeah, privacy is the elephant in the room. In Europe, GDPR requires that employees be informed if their data is used for automated decision-making. But in the US, there's no federal law yet. A few states, like California and Illinois, have laws that apply, but it's patchwork. Luna: And even if it's legal, there's the question of trust. If employees know the AI is watching everything they do to feed into reviews, does that change how they behave? Lucas: Absolutely. And that's the paradox. The same technology that can make reviews fairer and more consistent can also make employees feel surveilled and less willing to take risks. It's a design challenge as much as a technical one. Luna: So where do you see this going in the next two years? Are we heading toward fully automated reviews? Lucas: I don't think we'll ever get to fully automated, because the human element is still crucial for nuance and empathy. But I do think within two years, it'll be standard for every large company to use AI to at least draft a first pass of the review. The manager then edits it, adds their own voice, and makes the final call. Luna: So the AI becomes like a research assistant, not the judge. Lucas: Exactly. And the best implementations are the ones where the AI is transparent about what it knows and doesn't know. If it suggests something based on data, it says so. If it's unsure, it asks for input. That kind of collaborative design could actually make reviews more human, not less. Luna: That's a nice note to end on. I think as long as the technology stays in service of the conversation — not replacing it — we might be okay. Lucas: Yeah. And that's the conversation we'll keep tracking as the tools evolve. Thanks for listening.