In a busy restaurant kitchen, the most important person on a Friday night often isn't cooking. They're standing at the pass, the counter between the line and the dining room, holding every ticket that's come in. They know table twelve has an allergy, that the fish for table four is two minutes behind, and that the server who just walked up is about to pick up the wrong plate. They call the orders, they check every dish before it leaves, and they send back anything that isn't right. The chefs get to cook because somebody else is holding the whole picture.
Nobody at the pass decides what's on the menu. That's the owner's call, and the chef's. The expediter's job is to make sure the people making decisions have what they need, when they need it, and that nothing leaves the kitchen wrong.
If you run a small insurance agency, you've probably noticed that nobody is standing at your pass. You are the principal, the top producer, the person who answers the carrier email about the missing signature, and the one who remembers which renewal needs a conversation before it needs a quote. You're cooking and expediting at the same time, and the tickets don't stop coming in because you're on the phone with a client.
What an AI Chief of Staff is in an agency
An AI Chief of Staff for an insurance agency is one AI assistant that holds the agency's full picture - your clients, your renewals, your carriers and your own priorities - and uses it to brief you, draft the busywork, and push back before you make a mistake, while every coverage decision and every word to a client stays with a licensed human.
That last clause is the one that matters most in this business, so I want it up front rather than buried. An AI Chief of Staff doesn't recommend coverage. It doesn't pick a plan for a Medicare client or tell a business owner which limits to carry. That's the advice people come to you for, and it's also the advice you're licensed and accountable for. What it does is get you to that conversation with the preparation already done.
The term gets used loosely, so here's the boundary in plain terms. A handful of AI tools, each holding a sliver of your agency's context, doesn't add up to one. Neither does a chatbot you open when an email needs rewriting, or an autopilot you've handed the keys. It's one seat with one job: hold the picture so your attention can go to the clients.
What it carries in a one-to-five-person shop
In a small agency, the work an AI Chief of Staff takes isn't glamorous, which is exactly why it's worth handing over. It's the work you do between the conversations that actually need you.
Monday morning is the easiest place to see it. Instead of opening the management system, the inbox, the calendar and three carrier portals to work out what the week holds, you start with a briefing: who you're meeting and what you promised each of them last time, which renewals are sixty days out and which of those have had a claim or a change since last year, and which emails from the weekend actually need you as opposed to your service person. You didn't assemble any of it. You read it with your coffee and decide where your week goes.
Through the week, it drafts. The recap letter after a review meeting, from your notes. The follow-up to the prospect who went quiet after the quote. The summary of what changed in a carrier's renewal terms, so you walk into the client conversation with the comparison already laid out. Everything it writes comes to you as a draft, and nothing goes to a client until someone who knows the client has read it.
And sometimes it asks you a question you'd rather not answer. The part of the role most people skip is the pushback: an assistant with standing instructions to challenge your assumptions will ask whether that discount you're about to offer one client is going to be a problem when three others hear about it, or whether you've actually got the capacity for the new book of business you're excited about. A good human chief of staff earns their keep in those moments. The AI version can too, but only if you tell it that's part of the job.
The difference between this and a chatbot
Plenty of agency owners have already tried ChatGPT or something like it and quietly concluded it's a clever toy. Given how they were using it, they're usually right.
A chatbot starts every conversation knowing nothing about you. You explain the client, the policy, what you already tried, what tone you need, and it does a decent job with that one task. Then you close the tab and it forgets all of it. Tomorrow, same client, you give the same speech. That rebuilding is its own version of the Toggle Tax, and it's why so many owners decide AI costs them more time than it saves.
A chief of staff keeps the context. It's been told who your agency serves, which carriers you write with, what your priorities are this quarter, and how you like to talk to clients. It has standing instructions: ask clarifying questions, challenge my assumptions, never send anything on my behalf. You can build one on the same AI platforms most people already pay for. What separates it from the chatbot is that somebody sat down and gave it a job description.
What it never does
In a regulated business, you write the rules before you switch the role on. Before an AI Chief of Staff touches anything real, write down the handful of things it never does, with the reason next to each one. In a Digital Crew that's a Flight Rule: a decision the AI never makes, decided in advance, with its reason attached so the person tempted to bend it at seven on a busy Tuesday night can see why it's there.
For most agencies, the short list looks something like this. It never recommends coverage, because that's the licensed judgment clients pay you for. Nothing it drafts reaches a client without a named person reading it first. Client information only goes into an AI account the agency has approved and set up properly, never a personal free account, which is where Shadow AI turns into an E&O problem. And when its summary disagrees with the policy document, the policy document wins.
Your carriers, your FMO if you have one, and your E&O carrier may all have their own requirements for how client data is handled, and theirs outrank anything you write here. Check before you connect a single system.
Where mine came from
I didn't set out to build an AI Chief of Staff. Toward the end of 2024 I'd set up a custom assistant in Gemini, named it CLU (it's Lloyd now), and given it information about me and my business. Then I drove from St. Louis to Los Angeles, and somewhere along the way I started talking through the coming year with it, and more to the point, why I felt like I was struggling.
I'd launched a podcast that May. People who heard it liked it, and it still wasn't getting traction, and I wasn't getting the interest or the business I thought I should. That was frustrating, and a long stretch of highway turned out to be a good place to say so.
By the end of the drive we'd agreed I needed a new website for my AI work instead of publishing it on my old marketing site, a target audience other than marketers, and a leaner way of running the business. Over the Christmas holidays, working with Lloyd the whole way, I chose the name, built out the audience and services, rebranded the podcast and built the site. It took two weeks. That business is The AI Hat, and Lloyd is still my AI Chief of Staff today.
What it was at the start was modest: a custom assistant with some information about me and a set of instructions, a long way short of what I'd tell you to build now. It didn't need to be sophisticated. It needed to know my business well enough to be worth talking to, and it needed permission to tell me things I didn't want to hear.
Who this is for, and who it isn't
An AI Chief of Staff fits best where one person is holding most of the picture: a solo agent, a principal with a service person or two, a small team where the owner is still the one everybody asks. That's where a single seat that knows the whole agency pays off fastest.
If you run a larger agency, with account managers, a service team and producers each carrying their own book, one chief of staff for the principal is a good start and not the answer. Your drag is spread across a dozen people, and it needs a Digital Crew: a Scout that finds and checks things, a Scribe that handles the notes and the follow-ups, a Coach that briefs the team before a meeting, each with a human who owns its output. The chief of staff is usually the first seat, because building it teaches you how the rest get built.
Where to start
The build itself takes about an afternoon, and I've written the whole thing up in How to Build an AI Chief of Staff: the context to load, the two instructions that matter more than anything else, how to connect it to where your work lives, and the rules to set before you do.
If you want one thing to do before you read it, open a blank document and describe your agency the way you'd brief a sharp new hire on their first morning. Who you serve, which carriers you write with, what a good week looks like, and the three things you'd most like to stop doing yourself. It's rarely written down anywhere. It's also the first thing your chief of staff will need, and it earns its keep whether you build one this month or not.
If you'd like to see where your team's hours are actually going before you change anything, take the AI Opportunity Score. It takes about three minutes, shows you how many hours a week your team could get back, and gets you a starter crew member built for the biggest one.
Somebody should be standing at your pass. It doesn't have to be you.
