I’ve been working in the tech industry for over 20 years, and for about half that time I’ve been in a leadership role. I’ve always believed that to achieve good results, it’s essential to keep an eye on teams’ wellbeing because how people feel shows up in their work.
People don’t always want to talk about what’s troubling them, so you need to spot it early when something’s gone wrong. I had a simple indicator: If someone started taking longer to get things done, I knew it was time for an unscheduled one-to-one. The drop in performance usually came down to overload or someone not having taken a holiday for a long time.
As AI has spread, this indicator has become less reliable. With AI tools taking over part of the workload, people can still hit their usual output even when they are a step away from burnout. This is not just my observation. In a study of 442 employees at tech companies, the link between burnout and a drop in productivity was weaker among those who used AI more.
So, if the dashboard says nothing about how the team is feeling, how can a manager spot the first signs of burnout in a team that is delivering excellent results?
THE NEW NORM
AI really does speed up work and save us time on routine tasks. Researchers at the University of Lausanne have even calculated exactly how much. They surveyed 83 managers and found that 82% of them save an average of 2 hours and 50 minutes per week thanks to generative AI. That’s about 1.5 working days a month.
You would expect people to spend this time on self-development or with their families, but no. More than half of the managers in the same study spent it on getting more work done, while 36% didn’t even realize that AI had saved them time.
AI doesn’t just help you take on more, but it also gradually raises the bar for what counts as a normal workload—and level of complexity—for a particular role.
An architect on my team once told me that since AI came along, he expects junior developers to be more independent. If they have questions, they should work it out with the AI assistant first. Then they can come back with a specific problem and possible solutions, rather than expecting a lecture on the basics of the profession. Another example: In an HBR article, the head of workforce innovation at KPMG says she now expects more of associates—they should come to meetings as well prepared as managers once did, because AI makes this possible.
These are positive examples of how AI can speed up growth for people in junior roles: They quickly become more independent and take on tasks that used to need more help from senior colleagues. But within a team, the productivity boost from AI is not always spread evenly. And that has consequences.
EXTRA WORK FOR SOMEONE ELSE
I recently came across an article by Sreenivasa Reddy, a lead software engineer at the telecoms company AT&T. In it, he described his team’s story, which initially looked like a textbook example of AI adoption—they started working 40% faster. But over the next two months, three of their strongest mid-level engineers resigned.
As it turned out, while junior and senior engineers were becoming more productive, the mid-level engineers were spending more and more time reviewing AI-generated code and fixing errors. The results of their work weren’t reflected in the metrics or performance reviews, and they received no credit for it, which, of course, was demotivating.
This isn’t just the story of one team. In a study by Workday involving 3,200 managers and employees across three regions, 85% said that AI saves them between 1 to 7 hours a week, but around 37% of that time is spent fixing and rewriting what it has generated. Often, this work falls to middle-level roles, as it did in Sreenivasa Reddy’s story.
HOW TO SPOT BURNOUT BEHIND HIGH METRICS
In the past, a drop in productivity might have been a sign to check in on someone. Now, it may come too late or never show up at all. So as AI speeds up the pace of work, I would advise leaders to look closer at what lies behind the rising productivity numbers. Here’s where to start:
- Acknowledge and value “invisible” work. A big part of working with AI barely shows up in productivity stats. It’s still routine work—reviewing generated content and fixing mistakes. The people doing this might have fewer completed tasks to their name, but they are the ones who keep the whole team’s numbers high. This kind of work should count both when balancing the workload and during performance reviews.
- Don’t turn every saved hour into more work. It is always tempting to finish one job quickly and move straight into the next. But when this becomes the norm, the team ends up operating on its limit. If AI frees up time, it is better to decide in advance how to use it. That might mean more work, but it could just as easily be quality control, learning, or simply taking a beat to recharge.
- Don’t confuse using AI with efficiency. For me, it’s about whether the tool helped us make a better decision. Otherwise, AI adoption just becomes another metric people optimize for the sake of optimizing.
I no longer take strong numbers as proof that the team is doing fine. After asking “How much have we achieved?” I ask one more question: “What did this result cost the team?”
Illia Smoliienko is founder and engineering leader at Sivense.
