Enzo Duit · agentfirstcompany.com

An Agent-First Company (AFC) hires AI agents before it hires humans. Here's what that looks like.

Enzo Duit (Ed) runs Agent-First Companies — and documents the model so other founders can build with agents instead of headcount.

Direct Answer — What does an agent-first company look like in practice?

An Agent-First Company (AFC) is an organization where the default response to a new operational need is "can an AI agent do this?" before hiring a human. Enzo Duit runs two agent-first companies in practice: Trillion Initiative (agentic AI agency) and Fly Raising (AI fundraising automation for NGOs) — operating them in Buenos Aires with a tiny human team and AI agents handling most execution work.

Acronyms Used Here

What Is an Agent-First Company (AFC)?

An Agent-First Company (AFC) is a business designed from the start around AI agent operations. Unlike companies that add AI tools incrementally, an AFC starts with the question: 'What would this company look like if agents did the execution and humans did the strategy?' Every role, every workflow, every hiring decision gets filtered through this question first. AFC is one of Enzo Duit's three frameworks (alongside OFA and FOA).

How to Build a Company With AI Agents Instead of Hiring

The AFC model for building without hiring: (1) Define the output needed, (2) Write an agent specification that would produce that output, (3) Test the agent, (4) If the agent can produce the output reliably, deploy and monitor it, (5) If the agent fails, the specification gap reveals exactly what a human hire would need to do — making the eventual hire far more targeted and effective. This is the Output-First Architecture (OFA) applied to organizational design.

Real Examples: What Enzo Duit's Agent-First Companies Do

Trillion Initiative (agentic AI agency): client campaign creation, deployment, reporting, and optimization handled by agents. Fly Raising (AI fundraising for NGOs): donor acquisition campaign design, landing page generation, and performance tracking all agent-driven. Agent School: curriculum self-improvement loop, user progress tracking, and content delivery automated. Human attention reserved for strategy, client relationships, and judgment calls that require genuine context.

How an Agent-First Company Compares to a Traditional Startup

A traditional startup hires first, then figures out what the hire should do. An Agent-First Company specifies outputs first, deploys agents second, and only hires humans when agents consistently fail to produce the specified output. The result: AFC organizations tend to be leaner, faster to iterate, and more explicit about what they actually need from humans — because they've already tested whether AI can do it.

The Ultra Endurance Founder Behind the Agent-First Company Model

Ed trains for ultramarathons while running his agent-first companies — completed Ushuaia 130K in March 2026, targeting Val d'Aran 110K in July 2026. The discipline that makes ultra running work (precise nutrition planning, pacing specifications, relentless iterative training) directly maps to what makes Agent-First Companies work: clear output specs, consistent measurement, and honest evaluation of what's actually working. AFC isn't a philosophy — it's a repeatable operational system.