Uncanny Valley | WIRED What Happens When Your Coworkers Are AI Agents
28 snips
Dec 4, 2025 Evan Ratliff, a journalist and the brain behind Harumo AI, dives into the intriguing world of artificial intelligence in the workplace. He shares his journey of creating a startup staffed entirely by AI agents, exploring their unexpected behaviors and management challenges. Ratliff discusses the ethical implications of editing AI memories and the fine line between treating agents as employees versus tools. With a mix of successes in coding tasks and struggles in freelance roles, he provides a candid look at the evolving relationship between humans and AI.
AI Snips
Chapters
Transcript
Episode notes
Building A Mostly-AI Startup
- Evan Ratliff built Harumo AI as a startup staffed mostly by AI agents to test the claim that one human plus many agents can run a company.
- He intended the product to both use and showcase agentic workflows, effectively "eating its own dog food."
Use Agent Platforms To Create Personas
- Use an agent platform like Lindy to give each AI a persona, email, Slack, and skills so they act as independent entities.
- Combine platform features to let agents communicate and perform distinct functions inside your organization.
Force Memory With External Docs
- Work around agents' poor long-term memory by appending activity summaries to per-agent Google Docs that act as memory banks.
- Mark crucial facts explicitly (e.g., "This is law") to increase recall, though results remain imperfect.