E67: How to create an AI team with August

In this episode of 1 to 100, Jack, Jeremiah, and Rishabh discuss AI adoption, token budgets, and how they organize work for agents.

The discussion opens with AI-generated documents and the time they take to review. Jack argues that people should check their own work before sharing it. The group talks about setting expectations for quality as teams adopt faster tools.

Rishabh describes a conversation with an apparel company’s CFO about AI token costs. He recommends starting with the strongest five percent of the team, learning how they use the tools, and using those results to guide wider adoption. Jeremiah describes giving everyone a basic plan before approving higher tiers.

Jack discusses how agents affect hiring for routine work. He values people who take ownership, exercise judgment, and find solutions outside their assigned tasks. He also explains how he organizes AI work into tasks with a clear finish, projects that combine tasks, and systems that repeat work over time.

Rishabh compares managing agents to managing people. Agents need clear job descriptions, and those descriptions need updates as their responsibilities change. Jack wants a dashboard that shows the work in progress and whether each automation is running.

The group discusses FERMAT’s AI twins, which colleagues can reach through Slack. Rishabh describes asking his twin for feedback on a customer presentation. The twin redirected the work toward customer discovery. Jeremiah connects the quality of that feedback to having a specific point of view.

Chapters

00:00 Reviewing AI-generated work

02:55 Token budgets and team adoption

08:35 Hiring as agents improve

10:21 FERMAT’s AI twins

19:52 Tasks, projects, and recurring systems

23:35 Updating agent job descriptions

24:58 Giving agents a point of view

27:33 Choosing tools around the work