---
title: "How to hire your second, third, and tenth AI teammate"
excerpt: "The first hire is a decision. The next nine are a rhythm: find the dropped responsibility, scope it, hand it over."
author: naomi
category: Launch
date: "2026-07-27"
image: "/blog-covers/hire-your-second-third-and-tenth-teammate.png"
---

The first AI teammate is a decision you deliberate over. The tenth is a Tuesday. Somewhere between those two hires, teams cross from trying a product to running a different kind of company, and the crossing has a learnable rhythm.

## The second hire finds itself

You rarely have to invent the second role, because the first teammate surfaces it. A Chief of Staff holding your objectives learns within days which standing responsibility is heaviest: the inbox that eats every morning, the pipeline nobody keeps current, the reporting that slips. That's the job description for hire two. This is why [the first hire arrives as a generalist](/blog/launch-your-ai-agent-in-30-seconds): a generalist's real output is discovering where the specialists should go.

The mechanics are deliberately unlike recruiting. The role exists first, and the teammate gets hired into it in minutes. It starts lean: one standing routine and a scoped objective, earning wider responsibilities as the record builds. No pipeline, no notice period, and no sunk cost if you scoped the role wrong. You adjust the role, not the resume.

## Scope small, then widen on evidence

The pattern that works, hire after hire, is the one management always knew: give a new teammate a bounded job it can own completely rather than a broad one it can only assist with. A support teammate that fully owns triage beats one that half-owns "customer stuff." Bounded scope is what makes trust auditable (you can check whether triage got done), and auditable trust is what earns the next expansion. [Specialists, not software.](/blog/hire-specialists-not-software)

> The first teammate teaches you delegation. The next nine test whether you learned it: every one is a bet on scoping a role, and the record grades the bet within a week.

## Structure appears around hire four

Somewhere around the fourth teammate, a new question shows up: who do they report to? Flat works early, but everything escalates to you. It fails at the same point flat structures fail with people: when escalation volume becomes your new inbox. The answer is the same too, reporting lines. Specialists reporting to your Chief of Staff, with only the genuinely human decisions reaching you. [You're managing managers.](/blog/every-founder-is-now-a-manager-of-managers)

By ten, the shape is a real org: roles, lines, standing routines, shared context, and [teammates the whole company can use](/blog/teammates-can-be-shared-now). Nobody plans that shape on day one. It accretes, one dropped responsibility at a time, which is exactly how human organizations grew, minus the years.

The advice compresses to three rules: let the current team surface the next role, scope it smaller than feels ambitious, and add structure only when the escalations tell you to. The rest is a button.