Aditya Sharma, co-founder of Keel, which develops AI troubleshooting software for IT teams, frequently works through the night. His generative AI agents handle multistep operations—data extraction, code generation, customer feature development, research for proprietary models, marketing campaign monitoring.
The economics are stark: blocking agents for eight hours costs money. They can execute around the clock. So Sharma does too.
"They demand your attention," Sharma said. "Does it need anything? Can I help it in any way?"
Peter Pezaris, 56, launched Proxon, an AI workforce-management software company, four months ago—his fifth startup in 30 years. He had promised his wife he would retire this year. Instead, he works 7:30 a.m. to 2 a.m. daily.
Proxon, based in San Mateo, California, is projected to hit $1 million annual revenue run rate by October. Pezaris estimates six human developers plus AI agents operate 30 times faster than developers alone. He compared the experience to an intoxicant: "I've never actually done drugs but I imagine it's what it feels like."
Yet the productivity gain creates a compounding work problem. Faster onboarding generates higher customer request volume. Pezaris articulated the trap: "Every minute that I'm not working, I'm missing out on a week's worth of work."
Abby Grills, co-founder of Riveter, an AI-enabled data-gathering platform, reported the same dynamic—viewing every hour of work as potentially determinative for company survival.
Mandy Haskett, a leadership consultant, said founders with significant responsibilities struggle to deprioritize work, often pushing themselves to illness due to employee and investor reliance. Pezaris himself was hospitalized twice as a younger founder, once sleeping only one hour per night for a month before collapsing.
