By Craig S. Cody, CPA, Certified Tax Coach. Published July 2026.

Here's the answer up front, because I don't like burying it. The headlines say AI is coming for the jobs. On Main Street, and in a 10 to 50 person agency, the opposite is happening. Small business owners aren't cutting people with AI. They're using it to help the humans they already have serve more clients, respond faster, and make fewer mistakes. That's a capacity story, not a layoff story. Big companies can cut whole departments because big companies have bloat. A 20-person agency doesn't. So the smart move isn't firing the people you fought to hire. It's using AI to raise your ceiling, and measuring it with one number per project so you know a real tool from an expensive toy.

I've worked with a lot of agency owners over the years, and right now a fair number of them are reading the same scary headlines you are and wondering if they're supposed to be trimming staff. Before you run the corporate playbook in a shop your size, let me show you why it doesn't fit, and what to do instead.

What's Actually Happening With AI in Small Business?

Owners are using it to expand what their team can do, not to shrink the team. A recent piece in The Guardian by Gene Marks, who's covered small business for years, tells two stories that stuck with me.

The first is a guy who sells windows and doors. He put about $10,000 into an AI tool for his showroom. It listens to the conversation between the salesperson and the customer, then drafts the quote automatically for the salesperson to review and send. In his words: "It allows my salespeople to talk to more customers and spend less time doing paperwork. And it cuts down on errors." Read that again. More customers, fewer mistakes. Nobody got fired. His salespeople got their afternoons back and pointed them at more selling.

The second is an owner who connected an AI assistant to a folder full of her product manuals, spec sheets, and technical documents. Now her support team asks it any question and gets an answer in seconds instead of digging through PDFs, and her next move is to open that up to her customers directly. Same team, more output, faster answers.

That's the pattern almost nobody in the media is telling you about. Exhausted owners using technology to help scarce employees do more, make fewer errors, and serve more customers. Not mass layoffs.

Why Doesn't the "AI Replaces Workers" Story Fit a Small Agency?

Because the layoff math only works where there's slack to cut, and a lean agency doesn't have any.

Marks makes a point in that piece I've been making to agency owners for years, just in different words. Big companies can lay off thousands of people without blinking because big companies have bloat. Whole departments, layers of middle management, entire functions where you can automate a slice and quietly cut heads. That slack was always sitting there.

Now look at your agency. Twenty people, maybe twenty-five. Tell me which one of them is dead weight. You can't, because in a shop that size there isn't any. Everybody's pulling. Everybody's valuable. You probably struggled to find half of them, and you'd struggle to replace a single one.

So when you take the corporate AI playbook, "adopt it, then cut heads," and drop it into a 20-person agency, you're not trimming fat. There is no fat. You're cutting into muscle, and you're firing capacity you'll need the second the next retainer lands. The demographics don't get easier from here either. The workforce is shrinking, good people are hard to find, and the last thing a smart owner does is treat their scarce, hard-to-replace team as the thing to eliminate. You don't have a headcount problem. You have a ceiling problem, and AI is a ceiling tool, not a headcount tool.

What Should AI's First Job Be in a 10 to 50 Person Agency?

Its first job is to turn the team you already have into more capacity you can actually sell.

Sellable capacity means four specific things: faster response to your clients, cleaner delivery with fewer errors, more clients served by the same humans, and more of your senior people's time freed up for the high-value work clients happily pay a premium for. That's the windows-and-doors guy. That's the support-docs owner. They didn't ask "how many people can I remove?" They asked "how much more can this same team handle, and how much better?"

That's the question for you too. Not "who can AI replace," which is the corporate question and the wrong one for your size. The right question is "where is my team bottlenecked, and can AI widen that bottleneck so we can take on more without breaking?" Capacity you can sell shows up in your revenue and your gross profit. A seat you cut in a lean shop just shows up as a hole the next time you win business.

Should a Human Still Review What the AI Produces?

Yes, always, and that review step is a feature, not a delay. It's the whole reason it's safe to use.

Go back to the windows guy. The AI drafts the quote. The salesperson reviews it and sends it. It doesn't fire the quote off on its own. Your clients feel this too. Marks points out that small business owners don't fully trust AI yet, and honestly, they're right not to. They've read about the breaches. They wonder whether their pricing and proprietary information are being exposed. They've been burned by buggy technology before, and they're not impressed when a vendor calls a flat-out error a "hallucination."

So nobody sensible is handing a bot the client relationship, the invoicing, or the final deliverable with no oversight. AI drafts, a human decides. In a professional services business, that judgment layer, the part where an experienced human looks at the output and says "yes, send it" or "no, fix this," is not the part you're trying to remove. It's the part clients are paying you for. AI makes the draft cheap. It makes your judgment more valuable, not less.

How Do You Keep "Capacity" From Becoming a Feel-Good Word?

You make every AI project name one operating metric it will move before you start. One. If it can't, it's a toy, not a tool.

"We feel more productive" is not a result, and capacity can turn into a fuzzy, feel-good word fast. So here's the discipline that keeps it honest. Before you buy the subscription, before you turn it loose on the team, the project has to name a single operating metric it's going to move, and it has to be one of these five:

• Cycle time so you shorten how long a project takes from kickoff to client approval, not just how fast one draft comes out.
• Rework so you cut how often something has to be redone because it was wrong the first time.
• Response time so a client hears back from you in an hour instead of a day.
• Capacity sold so the same team takes on more billable work without new hires.
• Gross margin so you keep more of what you bill after the cost of delivering the work.

That's it. One project, one number, named out loud before you begin. Here's what the rule does for you: it kills the toys. If someone wants to roll out a shiny AI tool and can't tell you which one of those five it moves, you have your answer. It's a monthly charge and a demo that impressed somebody. Pass. But if they can say "this cuts our first-draft cycle time in half" or "this drops our support response from a day to an hour," now you've got a real project, and in ninety days you can check whether it actually did what it promised. Every real win in that Guardian piece had a number attached. Every one.

A quick note so I don't muddy two ideas. This one-metric rule is a gate for each individual project: pick a number, prove that project earned its keep. Measuring whether AI moved the whole agency's profitability is a bigger question with its own scoreboard, and I wrote about the five numbers that tell you if AI is actually paying off separately. Start with the gate. It's the fastest way to stop paying for toys.

Why Is a CPA Telling You How to Think About AI and Staffing?

Because the cost of getting this wrong hides in exactly the places most accountants never look.

Most accountants look in the rearview mirror. They tell you what already happened. If you panic-cut two people this year because a headline told you to, your accountant will happily book the payroll savings and never once ask whether you just kneecapped your own capacity heading into your busy season. That's a windshield question, and historians don't ask it. On the flip side, "we used AI to serve 30% more clients with the same team" doesn't announce itself on a tax return either. It shows up quietly, in your margin and in your gross income per person, and you have to be looking forward to see it. That's the heart of running the agency by the numbers.

When you've worked with a lot of agencies over a lot of years, and I've been doing this more than 23, you learn to be suspicious of whatever the headline is screaming. Right now the headline is "AI is cutting jobs," and for shops your size it's mostly wrong. Corporate America might use AI to cut. You're going to use it to build, to help your team stop just surviving and start growing. That's a choice, and it's a numbers decision, not a headline decision.

Who This Is For, and Who It Isn't

If you run a 10 to 50 person agency, you're seeing the layoff headlines, and some quiet part of you is wondering whether you're behind for not trimming staff, this is for you. You're not behind. You're just running a different kind of business than the giants making those headlines, and the capacity play beats the headcount play almost every time at your size.

If you run a bloated operation with real slack to cut, some of this won't apply the same way, and that's fine. And if you're hoping I'll tell you AI is magic and you should replace your team to juice this quarter's payroll line, I'm not your guy. The agencies that flourish treat AI like any other investment. They point it at their ceiling, keep the humans who make the work good, and measure every project by one honest number.

Frequently Asked Questions

Is AI replacing jobs at small businesses and agencies?
Mostly no, at least not at small companies. Reporting from The Guardian and payroll data both point the same way: small businesses are using AI to help their existing teams serve more customers and make fewer errors, not to cut staff. Big companies can lay off at scale because they have departmental bloat to automate away. A lean 10 to 50 person agency doesn't have that slack, so cutting heads usually means cutting capacity you'll need.

How should a small agency use AI, then?
Use it to create sellable capacity from the team you already have. That means faster client response, cleaner delivery with fewer errors, more clients served by the same people, and freeing your senior humans for the high-value work clients pay a premium for. The question to ask is "where is my team bottlenecked, and can AI widen it," not "who can AI replace."

Should AI-generated work still be reviewed by a person?
Yes. In every solid example, the AI drafts and a human reviews and approves before anything goes out. That judgment layer is what clients are actually paying for, and it's also what keeps you safe given real concerns about data privacy and AI errors. AI makes the draft cheap; it makes your judgment more valuable, not less.

How do I know if an AI tool is worth paying for?
Make it name one operating metric it will move before you start: cycle time, rework, response time, capacity sold, or gross margin. Pick one, then check in about ninety days whether it actually moved. If a tool can't name a number it's supposed to improve, it's a toy, not a tool, and you can pass.

What's the difference between this and measuring AI's overall ROI?
This is a per-project gate: each AI project picks one number and proves it earned its keep. Measuring whether AI raised the whole agency's profitability is a broader question with its own scoreboard (AGI per employee, realization, cycle time, rework, and capacity converted to revenue). Start with the one-metric gate on each project, then zoom out to the full scoreboard once you're past the toys.

Let's Talk

Want a second set of eyes on that list? Take every AI tool your agency is already paying for, or thinking about, and write the single operating metric each one is supposed to move. The tools with a real number next to them are worth keeping. The blank spaces are toys you're funding. A free tax and profit analysis is where we look at the real numbers together, capacity, margin, and what AI is actually doing to your bottom line, and I show you where it's helping and where it's leaking. Let's talk.

Craig S. Cody is a CPA, Certified Tax Coach, and former NYPD Lieutenant who helps agency owners keep more of what they make through proactive, year-round tax planning and fractional CFO work. His firm works with a large number of Agency Management Institute members. The small-business AI examples cited here come from Gene Marks's reporting in The Guardian (July 26, 2026); the agency interpretation is my own.