The AI Operating Model in Practice

Building an AI agent is not the same as building an organization that can operate AI systems. This series follows the technical, operational and organizational decisions needed to move from experiments to reliable production: permissions, product ownership, behavior tests and controlled releases.

Start with Part 1

What you'll learn

Check permissions before a tool runs and again where the business action happens.
Give product teams ownership of the agents they ship, including unresolved outcomes after deployment.
Turn agent failures into tests that check proposed actions, service requests and business outcomes.
Treat behavior changes as releases, with qualified review capacity and someone who can stop new work.

Read in order

Part 1 · Start here
The AI Operating Model in Practice - Enforce agent permissions before a tool runs - Part 1
An agent can propose a refund but it cannot grant itself permission. Check order access before a tool runs and repeat the check in the business service before a refund attempt.
Sep 28, 202612 min read
Part 2
The AI Operating Model in Practice - Give product teams ownership of the agents they ship - Part 2
A refund assistant can pass permission checks and still leave customers with wrong proposals or unresolved payments. Give the product team an owner who can stop intake, fix the workflow and see every affected case through to an answer.
Oct 1, 202612 min read
Part 3
The AI Operating Model in Practice - Turn agent failures into release-blocking tests - Part 3
A refund assistant can give a convincing answer while requesting the wrong amount. Check its proposed action, attempted service requests and stored business outcome before shipping a change.
Oct 4, 202612 min read
Part 4
The AI Operating Model in Practice - Treat every agent behavior change as a release - Part 4
A prompt change can send more refunds to specialists without changing code. Decide its release using tested actions, qualified reviewer time and someone who can stop new admissions.
Oct 8, 202612 min read

Make human review capacity an AI rollout gateAn AI workflow can pass its quality checks while leaving too little time for human review. Use a review worksheet and a launch rule that checks qualified staff, unfinished work and decision deadlines before admitting traffic.Treat timed-out agent writes as unknown, not failedA timed-out write may already be complete. Save its operation identity before sending it and keep uncertain outcomes pending. Recover the original receipt through downstream support rather than sending a new business action.Make agent tool retries idempotent before enabling automatic recoveryAn agent may retry a tool call after a timeout even though the first call finished. Give each action a stable key and save its result so a retry returns that result instead of creating a second ticket, payment, message, or deployment.

What you'll learn

Read in order

The AI Operating Model in Practice - Enforce agent permissions before a tool runs - Part 1

The AI Operating Model in Practice - Give product teams ownership of the agents they ship - Part 2

The AI Operating Model in Practice - Turn agent failures into release-blocking tests - Part 3

The AI Operating Model in Practice - Treat every agent behavior change as a release - Part 4

Related reading and code