A landscaping company owner in Burlington asked me a question last summer that I still think about.
We had just finished scoping a quoting-automation project for him. His estimator was spending around 18 hours a week building quotes from site visit notes, photos, and an outdated price book in a shared spreadsheet. The proposed system would read the notes, match line items against current pricing, and produce a draft quote his estimator could review and send. Realistic time saved: 12 to 14 hours a week.
He looked at the proposal and said, in a quiet voice, “So I’m laying her off.”
I asked him why he thought that.
He said because that’s what AI projects do. You pay the firm, you save the hours, you reduce the headcount. That’s the math.
The estimator had been with him eleven years. She was the person who answered the phone when his mother died. He was already, in his head, working out severance.
It took us about twenty minutes of conversation to land on the actual answer, which was that he had no intention of laying her off, did not want to lay her off, and in fact had a long list of work he had been wanting her to do for two years and had never had time to assign. The 12 to 14 hours she was about to get back were not the end of her job. They were the start of a different one.
This is the conversation I now have on almost every engagement, and most owners walk into it with the same default. AI saves hours, therefore AI eliminates people. That math is wrong most of the time in the businesses I work in.
Where the bad framing comes from
The “AI replaces workers” headline gets clicks. It is also the framing most public conversations about AI are organized around, so small business owners walk into their first engagement having absorbed a year of media that treats the whole thing as a layoff story.
Big enterprise sometimes is a layoff story. A 4,000-person customer support operation absorbing AI deflection does produce a smaller headcount. The shareholders demand it.
A 12-person landscaping company is not that. Neither is a 22-person accounting firm or an 8-person dental practice. The people inside those businesses are not interchangeable units of labor. They hold institutional knowledge that exists nowhere except in their head, and their loyalty is part of what the business is. Owners who lay off after a successful AI project usually regret it within six months, when they discover what those people were quietly holding together that nobody had written down.
What the hours are actually worth
When an AI project returns 12 hours a week to a senior employee, those hours have three possible destinations. Most owners only consider one of them.
The first is payroll reduction. Cut the role, save the salary. It produces the smallest long-term gain, because the person you are cutting almost always knows things you have not documented.
The second is capacity for work the business needs but has been delaying. Every small business has a list of things the owner has wanted to do for two years and has never had time to start. A real customer-retention program. A review of supplier contracts. A cleanup of the CRM nobody trusts anymore. These projects would generate real revenue or eliminate real cost, and they live on a list nobody gets to.
The 12 hours your senior employee just got back are exactly the resource that list has been waiting for. The estimator at the landscaping company spent her returned hours on customer follow-up work that drove a 22 percent increase in repeat business in the following six months. None of that was in the original project proposal. None of it would have happened if she had been laid off.
The third is growth absorbed by the same headcount. When AI shaves hours off existing work, the same staff can handle more volume without growing payroll. The dental practice we worked with last year did not lay off the front desk coordinator after we automated insurance verification. They added 40 percent more patient volume with the same team. Revenue up, nobody’s job gone.
The exception, named honestly
There is a version of this where the cut is the right answer. If an AI project automates the entire job of a role hired specifically to do that one thing, and there is no plausible reassignment to other work, the math does point at a layoff. I have seen it maybe twice in 30-something engagements. Both times the role was newer, hired in the last 18 months, with a narrow job description and no broader institutional knowledge. The owner gave generous notice and offered placement help. That is the right way to do it when it is genuinely the right call.
I am naming this because I do not want to pretend AI projects never produce layoffs at the small business level. They sometimes do. The point is that “sometimes” is much rarer than the media coverage implies.
If you are about to cut someone after an AI project, the question I would ask is: what work have you been wanting to do, that you have not had time to start, that this person could do instead? If you cannot name something, the cut may be honest. If you can, you are about to lose someone who knows your business for a savings that will look smaller in six months than it does today.
What this looks like in the planning phase
The owners who get this right do one thing differently before the project starts. They write down, in advance, what they intend to do with the hours the project returns. Not vaguely. With a specific assignment for a specific person to a specific second piece of work.
The estimator at the landscaping company had a written plan for her next 18 hours before the automation went into production. Customer follow-up program. Referral tracking. Training for the junior estimator the company had hired the year before. When the hours arrived, they had a destination. There was no awkward week where she was sitting at her desk with less to do and an obvious payroll question forming in the owner’s head.
This sounds too obvious to be worth writing down. In practice, almost no owner walks into an AI project with the redeployment plan already drafted. The proposal talks about hours saved and the conversation stops there. The work of figuring out what those hours become happens after the project has shipped, and by then the owner has had three weeks to think about it as a payroll line item rather than a planning question.
Why I think this matters
The honest version of the AI conversation at the SMB level is not that it eliminates people. It is that it changes what those people spend their week on. If the owner is paying attention, the change can be a gift. If the owner is not, it becomes a slow loss of institutional knowledge that made the business work in the first place.
I would rather have the awkward conversation about redeployment up front than the awkward conversation about severance six months later. Both for the people whose week is about to change, and for the owner who is about to find out what their long-tenured staff were quietly doing.
If you are weighing an AI project right now, the most useful homework is one question. If this project returns 12 hours a week to one of my senior people, do I have a real plan for what those hours become?
If the answer is “I don’t know yet,” do the thinking before the project starts. Not after.