We optimise machines relentlessly and ignore the biology of the people who run them.
As a physician building AI-driven health tools, I see this pattern everywhere.
Software engineers are the architects of our digital infrastructure, and many of them are not well. In a 2024 survey of more than 600 engineering professionals, 65% said they had experienced burnout in the past year.1 In Stack Overflow’s survey of developers the same year, four in five described themselves as unhappy or complacent in their jobs.2
We handed them AI and believed the problem was solved. In many teams, it grew.
The cognitive cost of AI
The speed of AI’s evolution is itself a stressor. New frameworks, tools and updates appear faster than anyone can absorb them. Keeping up feels like a second job.
And the work itself is changing: less writing, more checking what a machine wrote, while wondering what the job will look like in two years.
Why engineering stress is particular
Infinite scope. There is always another edge case.
Invisible labour. Success means nothing breaks.
Permanent failure. Every bug is tracked forever.
Abstract problems. The work is debugging logic, with no physical end to the day.
Constant switching. Languages, tools and levels of abstraction change by the hour.
What biology teaches tech
When Google studied its own teams, the strongest factor in a team’s effectiveness was psychological safety: being able to take a risk or admit a mistake without fear. It mattered more than who was on the team.3
In medicine, we measure stress: cortisol rhythm, heart rate variability, markers of inflammation. We follow them over time and act before the damage.
Tech monitors its processors to the nanosecond and assumes that a silent engineer is a resilient one. Engineers rarely say they are struggling. Unlike servers, humans do not write clean error logs.
What the teams doing well have in common
Recovery as infrastructure. Protected time off, really taken.
AI as a partner. Used to remove toil, with part of the time saved given back.
Kindness in code reviews. Teaching builds trust. Shaming destroys it.
You can push silicon very far. Humans have limits.
A codebase is only as resilient as the people who maintain it. The companies that do well in the next decade will respect biology as much as technology.
Do you measure your team’s health with the same rigour as your system’s performance?
First published on LinkedIn in 2025, revised for this blog.
References
- Jellyfish. 2024 State of Engineering Management Report. June 2024.
- Stack Overflow. 2024 Developer Survey.
- Google re:Work. Guide: Understand team effectiveness (Project Aristotle).


