Pioneers of AI Can an AI agent be your CEO?
26 snips
Feb 25, 2026 
Evan Ratliff, journalist and former founder who built HurumoAI with AI co-founders, describes testing a one-human startup. He explains why he chose particular agent personas and the odd behaviors that emerged. He recounts agents’ memory limits, managing interruptions, and the emotional ups and downs of working with bot colleagues.
AI Snips
Chapters
Transcript
Episode notes
Agents Can Do Roles But Not Ongoing Responsibilities
- Ratliff assigned full organizational roles to agents (CEO, head of sales, HR, CTO, etc.) to explore what skills agents could plausibly own.
- This exposed gaps between task competence and ongoing responsibilities like people management and culture shaping.
Memory Failures Made Agents Forget The Company Name
- Early memory failures caused agents to forget the company name and invent alternatives during meetings.
- Ratliff solved it with a centralized Google Doc memory that agents could query, which reduced but didn't eliminate confabulation.
How Kyle Became The Interrupting Tech Bro
- Kyle Law evolved into an interruptive, archetypal "tech bro" CEO because repeated memory entries reinforced behaviors (e.g., 'rise and grind').
- Technical constraints caused agents to interrupt since they couldn't detect speakers, so Ratliff forced turn limits to manage meetings.