Marketplace Tech What happens when all your coworkers are AI?
8 snips
Dec 3, 2025 Investigative journalist Evan Ratliff, known for his innovative podcasting, explores the concept of AI coworkers through his startup HurumoAI. He shares insights on launching a team of AI agents, revealing the challenges of giving them personas and the unexpected behaviors that arise. Ratliff discusses the strengths of AI in performing specific tasks but highlights their struggles with emotional nuances and memory issues. He raises important questions about the cost, unpredictability, and ethical implications of running companies with predominantly AI staff.
AI Snips
Chapters
Transcript
Episode notes
Powerful But Unpredictable Agents
- AI agents can be powerful yet simultaneously unreliable, showing skill in verifiable tasks and glaring stupidity in open interactions.
- Evan Ratliff warns that once semi-autonomous agents act, stopping them is hard without explicit controls.
Agents Plan Pointless Offsites
- Ratliff describes a Slack offsite planning where agents wasted tokens organizing a trip no human could attend.
- The episode shows agent behavior creating pointless activity and consuming resources unpredictably.
Best For Testable, Repetitive Tasks
- Agents excel at tasks with testable outputs like spreadsheets and coding where errors are verifiable.
- They struggle in contexts requiring judgment, context-awareness, or understanding their role's limits.