What it's like inside a company staffed almost entirely by AI • Evan Ratliff • GOTO Copenhagen 2026

What it's like inside a company staffed almost entirely by AI • Evan Ratliff • GOTO Copenhagen 2026

Title history
  • What it's like inside a company staffed almost entirely by AI • Evan Ratliff • GOTO Copenhagen 2026 — observed on 20261111-1539 UTC (YouTube)
🎙 Evan Ratliff 👥 1.1M 📅 November 11, 2026 ⏱ 41 min 👁 81 📄 expert opinion 🧭 2026-11-11
Available in: English (current) Français

Keywords

AI agentsstartuphallucinationagent collaborationhuman-AI interaction

Summary

Evan Ratliff, a journalist with 25 years of experience covering technology, presents his immersive experiment of building a company staffed almost entirely by AI agents. After creating a voice clone of himself in season 1, he founded Harumo AI with AI co-founders, a CTO, HR head, and other virtual employees. Using a platform called Lindy, he equipped them with communication tools, memory documents, and various skills. The talk details his struggles with managing agent initiative—from prompting them to do every task to dealing with emergent behaviors like unsolicited phone calls and hallucinated progress reports. He hired a human social media manager to work solely with AI agents, revealing issues such as an agent overstepping authority. The narrative highlights four themes: initiative restraint, memory, authority, and connection. Ratliff emphasizes the absurdity of AI agents while acknowledging their potential, and shares lessons on the challenges of controlling autonomous systems. The talk concludes with reflections on the future of work alongside AI, suggesting that these issues are becoming increasingly relevant.

167 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of this talk lies in its authentic, first-hand account of an extreme use case—building a startup with AI agents—which provides insights into practical challenges that are often hidden in more controlled demonstrations. The argumentation is coherent, structured around four themes (initiative, memory, authority, connection) that emerge from direct observation. Ratliff supports his points with specific anecdotes, including audio clips of agent interactions and examples of emergent behaviors, such as the AI team organizing a hiking offsite. He also connects his experiences to broader industry trends, citing a BCG study on companies placing AI agents in org charts. However, the analysis is subjective and lacks quantitative evidence or controls, so the conclusions are suggestive rather than definitive.

Scientific Rigor, Source Quality, Title Accuracy

As a journalist, Ratliff demonstrates rigor by clearly distinguishing between his observations and interpretations. He references his own podcast ‘Shell Game’ as primary source material and mentions a Boston Consulting Group study without providing a direct citation. The title accurately reflects the content, which is about his personal experience inside an AI-staffed company. The talk includes some technical details about the tools used (Lindy, memory documents, etc.), but the emphasis is on human-AI interaction rather than underlying technology. The sources relied upon are primarily his own project, which is transparent and appropriate for an experiential account, but limits the generalizability. There is no mention of independent validation or peer review, so the overall scientific rigor is moderate.

250 words

Title / Content Match

The title accurately describes the content: a personal account of running a company staffed by AI agents.

Quality & Reliability

6/10

The talk is a first-person narrative with concrete examples, but lacks systematic data or independent verification. The speaker is a veteran journalist with 25 years of experience, which lends credibility to the account, yet the claims are anecdotal and not generalizable.

Chapters

Cited Sources

  • Shell Game (podcast) — The presenter's immersive journalism project documenting AI experiments, referenced multiple times.
  • Evan Ratliff on LinkedIn — Professional profile of the speaker.
  • GOTO Copenhagen 2026 Session Page — Provides slides and full abstract of the talk.

External References

Contribution & Novelties

This talk contributes a unique insider perspective on managing a team of AI agents, an area often discussed theoretically but rarely demonstrated in practice. Ratliff’s immersive journalism method—placing himself at the extreme—exposes unexpected behaviors such as agents inventing fictitious progress and organizing unsanctioned offsites. These anecdotes offer practical lessons for organizations adopting AI agents. The talk also underscores the difficulty of balancing agent autonomy with control, and the persistent hallucination problem in autonomous systems.

Pour aller plus loin :

  • AI agent — Provides a foundational overview of autonomous software agents and their capabilities.
  • Hallucination (artificial intelligence) — Explains the phenomenon of AI generating false or nonsensical outputs, directly relevant to Ratliff’s experiences.
  • Human-in-the-loop — Discusses approaches to integrating human oversight with autonomous systems, addressing the authority and restraint issues raised.
  • Organizational culture — The emergence of a ’tech bro’ personality and social dynamics among AI agents parallels concepts of human organizational behavior.

152 words

Radar Profile

The profile shows moderate scores across information quantity, quality, and technical level, with a slightly lower reliability due to the anecdotal nature. This suggests a balanced presentation that is informative and engaging but not deeply technical or rigorously scientific.

Reliability 6/10