Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Is Reshaping Your Performance Review
Transcript
- Lucas: So there's this Fortune 500 retailer — I won't name them, but think big-box, thousands of locations — and last year they decided to scrap their annual performance review system entirely. No more 'meet with your manager in February, get a rating, hope for a raise'. Luna: That's a huge shift. What did they replace it with? Lucas: An AI system that essentially reads their internal communications — emails, Slack messages, project management updates — and generates ongoing performance feedback. It's not grading employees like a test. It's looking for patterns: who's collaborating effectively, who's solving problems, who's getting positive mentions from peers. Luna: So it's like a sentiment analysis of your daily work life. But doesn't that raise some pretty serious privacy flags? Lucas: Absolutely. And we'll get to that. But the retailer reported that after one year, employee satisfaction with the feedback process jumped 22 percent. Managers said they felt more confident giving feedback because the AI surfaced specific examples — 'You helped Maria on the Q3 inventory project' — instead of vague impressions. Luna: I can see how that would be more actionable. My old company did 360 reviews, and it was always a scramble to remember what anyone did six months ago. Lucas: Right. That's the core argument for ai driven reviews: recency bias and memory limitations are human flaws. An AI can track every interaction, every document edit, every kudos in a channel. The idea is to replace the once-a-year judgment with a continuous stream of data points. Luna: But 'every interaction' — that's a lot. Who decides which data matters? And what about context? Like, if I'm sarcastic in Slack, does the AI think I'm combative? Lucas: That is exactly the problem. Natural language processing has gotten better, but it still struggles with irony, sarcasm, cultural differences. One study found that NLP tools misclassify statements from Black employees as 'negative' at a significantly higher rate than those from white employees. So you could be introducing racial bias into performance evaluations. Luna: Wow. So the same tool that claims to reduce bias might actually encode it. What do companies do about that? Lucas: Some are building in what they call 'fairness audits' — they test the model on different demographic groups before deployment. Others limit the AI to only analyzing work output, not interpersonal communication. But the retailer I mentioned went all in: they analyze everything except direct messages, which they consider private. Luna: Still, even public channels — that's a lot of surveillance. Do employees get to opt out? Lucas: In this case, participation was mandatory. The company argued that since the data was already being generated on their systems, using it for feedback was a natural extension. But a survey of their employees found that 45 percent felt 'uncomfortable' with the level of monitoring, even if they liked the feedback quality. Luna: So there's a trade-off: better feedback for less privacy. Is that a dealbreaker? Lucas: It depends on the culture. Some tech companies, especially fully remote ones, have embraced this model. GitLab, for example, uses a lot of asynchronous text-based communication, so they've experimented with AI tools that summarize performance from merge requests and issue comments. They frame it as transparency, not surveillance. Luna: But GitLab is a developer-centric company. How does this translate to, say, a retail floor worker or a warehouse associate? Lucas: That's where it gets tricky. For deskless workers, the data sources are different — maybe time spent on tasks, customer ratings, security footage. Amazon has been criticized for using AI to track warehouse productivity and automatically fire underperformers. That's a much higher-stakes version of the same idea. Luna: Right. So we're not just talking about better reviews. We're talking about AI that can terminate your employment based on data you didn't even know was being collected. Lucas: Exactly. And that raises a legal question: in the U.S., most employment is at-will, so companies can fire for almost any reason. But if the reason is an AI algorithm, and the algorithm is wrong — or biased — the employee has very little recourse. There's no 'appeal the machine' process in most HR departments. Luna: Are any regulators looking at this? I know the EU's AI Act has provisions for high-risk systems, but does performance review software qualify? Lucas: It does, under the EU AI Act. Systems used for 'evaluation of workers' are considered high-risk, so they require conformity assessments, human oversight, and transparency. But in the U.S., there's no equivalent yet. The EEOC has started looking into algorithmic bias in hiring, but performance management is still a gray area. Luna: So companies that adopt these tools are essentially writing their own rules. That's a lot of power. Lucas: It is. And some are trying to use that power responsibly. There's a startup called BetterWorks that offers a platform where the AI surfaces feedback but only managers see the raw data — employees see a summary. Others like 15Five let employees set their own goals and the AI checks progress, not sentiment. Luna: I like that more. It feels less like Big Brother and more like a coach. Lucas: Right. And that's the distinction: is the AI a tool to help you improve, or a tool to judge you? The difference often comes down to transparency. If employees know what data is collected, how it's used, and have a say in the process, the acceptance rate goes way up. Luna: What's the actual data on that? Do people perform better under these systems? Lucas: A 2025 study from MIT and Harvard looked at a tech company that introduced an AI feedback tool. They found that teams using the tool had a 14 percent increase in productivity, measured by project completion rates. But they also found that the tool increased anxiety among employees who were already worried about job security. So the net effect is mixed. Luna: So it's not a silver bullet. It's a tool that has to be implemented thoughtfully. Lucas: Exactly. And here's where I think the conversation gets really interesting: the companies that are doing this well are not just deploying AI and walking away. They're combining AI with regular human check-ins. The AI provides data, the manager provides context and empathy. Luna: That actually sounds like the best of both worlds. But it also requires training managers to interpret AI output — which is a skill in itself. Lucas: Absolutely. And that's a cost that many companies underestimate. You can't just buy the software and expect it to work. You need change management, communication, and a culture that values feedback. Luna: Speaking of value — if today's conversation gave you something useful to think about, here's the thing. This show stays ad-free because of listener support. If you found this episode helpful, head over to buy me a coffee dot com slash fexingo. It's a simple way to keep these conversations going without any sponsors. Lucas: Yeah, we really appreciate that. It keeps us independent and focused on what matters to you. So back to the retailer — they're now two years in, and they're seeing some unexpected effects. Luna: Like what? Lucas: They found that the AI was surfacing a lot of 'negative' feedback about employees who were actually high performers — because those employees were the ones pushing for change, questioning processes, and sometimes being blunt. The AI flagged them as 'difficult' until the company adjusted the model to account for constructive criticism. Luna: That's a classic example of the AI missing context. A human manager would know that person is just passionate. Lucas: Right. And that's why the human element is still critical. But here's the thing: the AI also caught some subtle patterns that humans missed. For example, it noticed that one team consistently got lower feedback from cross-functional partners, even though their internal metrics were fine. It turned out the team was hoarding information — they weren't sharing updates with other departments. That was a blind spot for the manager. Luna: So the AI can see organizational dynamics that individuals can't. That's powerful. Lucas: It is. And I think that's the future: AI as a diagnostic tool, not a judge. It points out patterns, and humans decide what to do about them. Luna: I wonder how long until we see this in more traditional industries — like law firms or hospitals. Lucas: It's already happening in some. A few large law firms are using AI to track how associates contribute to case research and client communication. And in healthcare, there are pilot programs where AI analyzes patient outcomes and ties them back to team performance. But those are still early stage. Luna: So the genie is out of the bottle. The question is whether we can design these systems to be fair and transparent. Lucas: Exactly. And that's a design challenge as much as a technical one. The companies that get it right will probably have a competitive advantage in attracting and retaining talent. Luna: Especially as younger workers expect more frequent, data-driven feedback. Lucas: Right. Millennials and Gen Z have grown up with constant feedback from social media — likes, comments, ratings. So the annual review feels archaic to them. ai driven continuous feedback might actually be what they want, as long as it's done right. Luna: So maybe the cultural shift is already happening. The technology is just catching up. Lucas: I think so. But we need to be careful not to let the technology outpace the ethics. Because once these systems are in place, they're very hard to unwind. Luna: That's a good place to leave it. Thanks, Lucas. Lucas: Thanks, Luna. Talk to you next time.