Built AI Employees to Save Time? Research Shows You're Probably Working More

# Built AI Employees to Save Time? Research Shows You're Probably Working More

> **Quick answer:** Building AI employees or agents typically increases workload rather than reducing it through three documented mechanisms: task expansion (AI empowers non-experts to do your job badly, requiring your review), blurred boundaries (AI lowers friction so work bleeds into all hours), and multitasking overload (managing AI threads adds cognitive burden). A UC Berkeley HBR study from February 2026 documented all three, and a METR randomized controlled trial found developers actually worked 19% slower with AI despite feeling 20% faster. Your personality type determines how badly you fall into the trap.

If you just deployed AI employees to automate your workflow and you're somehow busier than before, you are not doing it wrong. You have uncovered one of the most documented and consistently ignored findings in the current AI boom. Researchers at Harvard Business Review and UC Berkeley spent eight months inside a 200-person tech company watching this happen in real time, and the results explain almost everything.

## The Study That Finally Explains It

In February 2026, Aruna Ranganathan and Xingqi Maggie Ye (UC Berkeley Haas) published an eight-month ethnographic study of AI adoption at a real tech company — not a survey, actual in-person observation. What they found killed the "AI saves time" narrative.

Three things happen the moment AI employees enter a workflow: