for the person who owns AI internally8 guides · ~67 min

Becoming an AI Agent Manager

Someone at your company is now responsible for making AI agents actually work — mapping workflows, wiring systems, writing evals, handling the team that won't adopt it, and proving ROI to the CEO. The title is still settling (AI Agent Manager, AI Ops Manager, Agent Operator all describe it), and most people doing the job were handed it with no playbook. This is that playbook: eight guides, in the order you actually hit them, written for someone doing it largely alone and not necessarily a developer.

The role: who owns AI agents inside a company

Aaron Levie predicts 500,000 to 1 million companies will hire someone to run their AI agents internally. The market is still arguing over the title — AI Agent Manager, AI Ops Manager, Agent Operator — but the job is real and most companies already have this person. What the role actually involves.

9 min

Your first 90 days — what to build, in what order

The sequence that avoids the most common failure: trying to automate everything at once. Which workflow to map first, how to ship one reliable agent before touching the next, and the milestones that prove the function is working by day 90.

10 min

Wiring your internal systems — without an engineer

What connecting Claude to your CRM, ERP, or ticketing system actually involves — what you can do yourself with connectors and what genuinely needs an engineer. The honest line between "configure it this afternoon" and "open a ticket with IT."

9 min

Evals without being a developer

The most-tested skill in agent-management interviews, and the one most operators skip. A no-code path to knowing whether your agent is actually reliable before someone in the company finds out it isn’t. How to build a test set you trust.

9 min

When your agent breaks — diagnose and fix

Agents fail differently than software. When the contract-review agent starts hallucinating clause details, where do you even look? A diagnostic process you can follow without reading code, and the fixes that actually hold.

8 min

Getting a resistant team to actually use it

The change-management problem nobody warned you about: the warehouse team routes around the agent, the finance team loves it. Why adoption stalls, and how to handle the employee who keeps going around the system because they don’t trust it.

8 min

Keeping costs under control as you scale

The monthly bill that starts small and quietly compounds. Where agent token costs actually come from, the threshold where you need to care, and the levers that cut spend without cutting capability.

7 min

Showing ROI to your CEO

What to measure and how to report it so the AI initiative survives the next budget review. The monthly update format that justifies your roadmap — and your own title and compensation review along with it.

7 min