Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / Why Your Office Sensors Know When You Quit
Transcript
- Lucas: So there's this thing happening in a handful of Fortune 500 offices right now where the building itself is effectively ratting out employees who are about to quit. Luna: The building? Like the HVAC system knows you're updating your LinkedIn? Lucas: Not far off. I'm talking about IoT sensors — badge swipes, desk occupancy sensors, even Wi-Fi connection logs. Companies are feeding that data into machine learning models to predict which employees are flight risks before they even talk to their manager. Luna: And this is actually working? Lucas: One retailer I looked at — a big national chain, name I can't share — they found that employees who stopped visiting the cafeteria for at least five consecutive days were 23 percent more likely to leave within the next month. And that pattern held even when you controlled for PTO and remote days. Luna: Wait — so skipping the cafeteria is a quitting signal? That's almost sad. They're eating alone at their desk before they even resign. Lucas: Or they're avoiding the chance run-ins where someone might ask 'hey, how are things?' because they don't want to lie yet. But the system doesn't care about the emotional subtext. It just sees a behavioral delta. Luna: Right. And if today's tech conversation gave you something usable, I want to mention that we keep this show ad-free by design. No sponsor segments, no mid-rolls trying to sell you an HR platform. If you find value in episodes like this one, you can support the show at buy me a coffee dot com slash fexingo. Lucas: Yeah, that direct listener support is what keeps us from having to pitch you anything. Appreciate anyone who chips in. So back to the sensors — the cafeteria pattern is just one signal. The bigger picture is that companies are now aggregating dozens of micro-behaviors into a single 'retention score' for each employee. Luna: Dozens? Give me a few more examples. Lucas: Drop in meeting attendance — especially recurring ones you used to never miss. Decline in internal messaging volume. Arriving later and leaving earlier even when you're in the office. One firm found that a sudden increase in printing personal documents correlated with exit within two weeks. Luna: Printing documents. That's so analog. But also, are these systems passively collecting all that data already, or are companies adding sensors specifically for this? Lucas: Most of it is repurposed. The badge system already logs every door entry. The Wi-Fi already tracks which floor you're on. The desk booking app already knows if you reserved a spot. Companies just started piping that data into HR analytics platforms that weren't originally designed for prediction. Luna: So the infrastructure was there, the use case just got repurposed. How many companies are actually doing this today? Lucas: The most cited number comes from a Gartner survey in late 2025 — about 22 percent of large employers with over 10,000 employees are actively using some form of workplace sensor data for retention modeling. That's up from roughly 8 percent two years earlier. Luna: That's a massive jump. And I'm guessing most employees don't know their cafeteria visits are being scored. Lucas: That's exactly the ethical knot. In the retailer example, they didn't tell employees the cafeteria data was being used for retention prediction. The policy said badge swipes were for security and occupancy optimization. Not for 'hey, you stopped going to the salad bar, we should worry.' Luna: So is there any legal protection here? I mean, in the US at least, employers can monitor pretty aggressively as long as they disclose it in the handbook. Lucas: That's the baseline. But the disclosure is often buried in a 30-page policy that says 'we may collect data for business purposes.' It's vague enough to cover almost anything. The EU is a different story — GDPR requires explicit consent for non-essential data processing. So a German automaker I spoke with had to get opt-in for their sensor-based retention pilot. Luna: And did people opt in? Lucas: About 64 percent did. But the pilot only worked because the company framed it as a 'wellness and support initiative' — the data would be used to offer early coaching or mental health resources, not to flag people for performance review. That framing matters a lot. Luna: So the same data, same model, but presented as support versus surveillance — and that changes adoption entirely. Lucas: Exactly. And the irony is the German pilot actually reduced turnover by about 11 percent in the test group. But if they'd run it as a surveillance program, they'd probably have increased distrust and maybe even driven people out faster. Luna: There's something else I've seen — some companies are combining sensor data with email metadata. So not just when you swipe in, but who you're emailing and how often. Lucas: Right. One tech firm in Seattle built a model that looked at the shrinking of an employee's internal email network — if you're emailing fewer colleagues over a four-week period, especially cross-department contacts, that was a stronger predictor than badge data alone. They called it 'collaboration decay.' Luna: Collaboration decay. That's a great term. And does it actually predict who quits, or just who's been assigned to a project they don't like? Lucas: That's the question. The models have decent signal to noise — the Seattle firm claimed 78 percent accuracy in predicting voluntary exits within 30 days. But false positives happen. Someone might reduce collaboration because they're heads-down on a deadline, not because they're job hunting. So the question is: do you act on every flag, or do you use it as a conversation starter? Luna: The best case is a manager getting a nudge: 'hey, maybe check in with Alex.' The worst case is an algorithm deciding Alex is flight risk and deprioritizing them for a promotion. Lucas: That's the risk. And some vendors in this space explicitly market their tool as a way to 'manage out' low performers who you suspect are about to leave anyway. That's a whole different use case from retention support. Luna: So there's a spectrum. At one end: proactive retention, check in with people. At the other end: preemptive replacement planning. Lucas: And in the middle, you have managers who don't really know how to interpret the data. They get a dashboard that says '3 of your 12 direct reports have elevated departure risk.' What do you do with that? Some managers have a good conversation. Others just start treating those employees differently — and the employee feels it. Luna: Which probably makes them more likely to leave. Self-fulfilling prophecy. Lucas: Exactly. One study from last year found that when employees were told their badge data was being used for retention analytics, trust in management dropped 14 points — even though the company had good intentions. The mere awareness of surveillance changed the relationship. Luna: So what's the right way to do this? If I'm an HR leader and I think we could reduce turnover with sensor data, how do I not mess it up? Lucas: The companies that do it well share three things. One: they're transparent about exactly what data is collected and how it's used — not buried in fine print. Two: they give employees the choice to opt out of the predictive model without penalty. Three: they only use the data for positive interventions — coaching, resources, manager training — never for disciplinary decisions or comp adjustments. Luna: And does that work? Do people still feel monitored even with those safeguards? Lucas: Some do. But the opt-out rate in the most transparent programs is around 12 percent — much lower than I'd expect. And the turnover reduction in those programs averages about 8 to 10 percent year-over-year. So there's clearly a business case. Luna: I wonder how this evolves as more companies adopt hybrid schedules. If someone only comes in twice a week, the sensor signal is much sparser. Lucas: That's the next frontier. Some vendors are now combining office sensor data with virtual collaboration data — Slack metadata, calendar patterns, document co-editing frequency. They're trying to build a 'digital presence' score that works even if someone rarely comes in. Luna: So the surveillance just moves from the physical to the digital. It's not less intrusive — it's just a different vector. Lucas: Right. And the same ethical questions apply. Maybe even more so, because digital data is so much richer — timestamps on every message, every document view, every meeting invite response. Luna: So where do you land personally? Do you think we're heading toward a world where every workplace interaction is scored for retention risk? Lucas: I think we're already there for about a fifth of large companies. The question isn't whether it will spread — it will. The question is whether the default frame becomes 'support' or 'surveillance.' And that's going to depend a lot on regulation, employee pushback, and how the early adopters tell their story. Luna: If I'm an employee listening to this and I'm worried my company might be doing this, what's the first thing I should check? Lucas: Look at your employee handbook's data privacy section — if it mentions 'occupancy analytics' or 'workplace utilization' or 'collaboration patterns,' there's a good chance they're already collecting the raw data. Then ask HR directly: is any of this data used in retention or performance models? In places with real transparency, they'll answer honestly. If they dodge, that's itself a signal. Luna: So the cafeteria might not be the only place you stop going. You might stop trusting the badge on your hip. Lucas: Exactly. And that trust loss might be the one metric no sensor can capture — until it shows up in the exit interview.