Why AI Belongs on Your Org Chart, Not in Your Tech Stack

• August 19, 2026

By Mike Allton

Recently, I ran an artificial intelligence (AI) training for a company just outside Pittsburgh with about 50 employees. We started with a simple exercise to inventory every AI tool people were already using on the job. There was no judgment; we just wanted to learn what’s really happening.

The list went on screen after screen. The CEO sat there stunned. Her staff had adopted AI throughout the agency, and leadership had no idea.

That company didn’t have an adoption problem. Adoption was going great. The problem was that nobody was in charge of it.

The 2026 Big “I” Agents Council for Technology Tech Trends Report says most independent insurance agencies are living some version of that reality. While 68% of agencies plan to increase AI use this year, 55% have no written AI use policy. A third of agencies describe themselves as “just experimenting,” while 8% say AI is embedded in their daily workflows.

Sit with that gap for a second, because the gap is the whole story. Experimenting with AI is easy. Building it into real workflows is rare. And what separates them, in my experience, has nothing to do with which tool you bought.

We treat AI like just another tech stack decision. Evaluate, purchase, distribute the login information, hope. That works fine for software. Your agency management system (AMS) certainly never needed a manager. But AI isn’t delivering features: it’s delivering work. The renewal summary, the client email, the proposal prep. And someone has to manage that work. That’s been the case in every business I’ve ever worked with.

When agency leaders ask me where to begin with AI adoption or implementation, I say something that sounds odd at first: Give AI a spot on your org chart. Write a job description that spells out what it handles and what it never, ever touches. Assign it a manager—an actual named human being who reviews what it produces—and write the policy. The 8% from the Tech Trends Report who achieved embedded status almost always have one, the 55% is the rest of us.

However, there is a wrinkle here. Boston Consulting Group’s Henderson Institute published a study in Harvard Business Review. In it, more than 1,200 managers reviewed the same error-filled document, and while some were told a person wrote it, others were told it came from an “AI employee.” The managers reviewing the AI employee’s work caught 18% fewer errors. Calling the AI a teammate actually made human oversight worse because people extended the same trust to the machine they’d give a colleague. As a result, accountability dropped.

However, the problem wasn’t putting AI on the org chart. It was putting it there with no manager attached. The study’s authors came to the same conclusion: Combine AI output with human review, on purpose, inside the workflow.

Here are three steps to incorporate AI with human oversight:

1) Name an owner. One person is accountable for AI across the agency. Usually, that’s an operations leader who knows the workflows cold, not whoever likes to play with tech the most.

2) Write the policy. Document which tools are approved, specify what client data never leaves the building and determine who reviews what.

3) Pick one workflow. Get it from experiment to embedded, with the owner checking output weekly, before you start working on the next workflow.

At that AI training in Pittsburgh, the shock wore off by the end of the afternoon, and it was replaced by relief. Because now there was a plan with actual employee names on it. Confidence showed up the minute somebody was in charge.

That’s really all this stage asks of us. Not more technology, more management. The next great hire at your agency might not be human, but it still needs a boss.

Mike Allton, founder of The AI Hat, helps independent agencies adopt governed AI. Take the AI Adoption Reality Check.