Can a small business actually afford an AI consultant?

In short: Affordability is not a revenue threshold. It is a volume question you can answer yourself with a calculator in ten minutes. Here is the arithmetic, four situations where the honest answer is do not buy, and what to spend instead if you are below the line.

Key takeaways

• You can test affordability yourself in ten minutes. Take the monthly fee, divide by the profit on one closed job, and ask whether you lost that many deals last month.
• Most small businesses cannot run that test, because the thing the system would fix is the thing nobody is currently measuring.
• Four situations where the answer is no: you cannot name the work, the volume is not there, nobody will own it, or the goal is to remove a person.
• If you are below the line, the correct spend is a hundred dollars a month and an afternoon, not eight thousand dollars and a consultant.

Start with the arithmetic, not the budget

Our voice agents cost about $1,000 a month once they are live, plus a setup fee. That is a real number, and it is the easiest one to test yourself. We publish [the rest of the price list too](/blog/what-ai-consulting-actually-costs).

Take your average closed job or signed lease and work out roughly what it puts in your pocket after costs. Say it is $500. At $1,000 a month, the system pays for itself if it produces two extra closed jobs a month. Not two extra calls. Two closed.

Now answer the actual question: can you name two deals last month that you lost because nobody picked up the phone?

If you can name them without thinking hard, the arithmetic already works and you have been carrying that cost quietly for a while. If you genuinely cannot think of one, do not buy a voice agent. That is not a sales technique. It is the same test we run on the discovery call, and it disqualifies people.

Run the same shape of calculation on whatever you are considering. What does the fee cost, what does one unit of the thing it produces earn, and how many units do you need. If you cannot make that arithmetic close, the answer is no, and any firm that talks you past it is selling you something.

The reason most people cannot run that test

Because the number is not in your books.

Missed calls are the clearest example. A call that goes to voicemail and never gets returned does not appear anywhere in your accounting. There is no line item called leads that evaporated on Saturday afternoon. The business has a rough feeling that it happens, and no idea of the size.

When we instrumented the phone lines for a property management operation and its brokerage arm, we pulled a clean sample of 277 real calls. Twenty seven percent of them came in outside business hours or on the weekend. Nearly one in four. Before the system existed, nobody in that business could have told you that number, because nothing was recording it. On the brokerage line, a routable callback was captured from about seven of every ten calls, and two thirds of the property management calls were resolved without a human touching them at all.

The first return on that build was not efficiency. It was finding out how big the leak was. You cannot compute a return on something you were not counting, and for most small businesses the honest first step is measurement, not automation.

If your test above came back "I don't know," that is a real answer and it points somewhere useful. Spend a month writing down every call you miss and every quote that goes out late. Then run the arithmetic again with a real number in it.

Four times the answer is no

You cannot name the work in one sentence. If you cannot finish the sentence "every week, someone here spends hours doing ______," there is nothing to build yet. Requests that arrive as "we should be using AI" are not requests, they are anxiety. There is nothing wrong with the anxiety. It is just not a scope. The [five questions an audit answers](/blog/five-questions-every-audit-answers) are mostly about turning that anxiety into a sentence.

The volume is not there. Automation earns its keep on repetition. If the task happens four times a month, a person doing it four times a month is the correct system, and a $12,000 build to remove it is a worse business decision than leaving it alone. The math on a five person shop is genuinely different from the math on a fifty person shop, and pretending otherwise is how firms in this category end up selling people things they regret.

Nobody inside the business will own it. Every deployment needs a named person whose job includes noticing when the thing is behaving strangely. Not the IT guy who has access. A person with time on their calendar. We have watched good builds die quietly because no operator inherited them, which is most of [why AI pilots die after the demo](/blog/why-most-ai-pilots-die-after-the-demo). If the answer to "who owns this on a Tuesday morning" is "we'll figure that out closer to launch," you are not ready, and the money will be wasted regardless of how well the thing works.

The goal is to remove a person. We do not build systems whose purpose is to cut headcount. That is a real refusal that costs us real deals, and we would rather say it here than three meetings in. If the business case you are running requires a salary to disappear for the numbers to work, we are the wrong firm and you should know that before you spend a call on us.

If you are below the line

You are not stuck. You are just not a consulting purchase yet, and there is a version of this that costs about a hundred dollars a month.

Buy your team proper seats on a good AI product rather than everyone using free accounts on their phones. Pick the single most repetitive thing in the business and rebuild just that one with an off the shelf automation tool. Give it a month. Watch what breaks.

Two things come out of that. Either it works, in which case you have solved the problem for a hundred dollars and you should keep the eight thousand. Or it half works, and the reasons it half works are precisely the specification for what you would eventually hire someone to build. Either outcome is worth more than a proposal.

This is the same reason we tell people the person who should be leading this work is usually [already on the payroll](/blog/your-most-important-ai-hire-is-already-in-the-building). The bottleneck in a small business is rarely the technology budget. It is that nobody has been given the room to look at the operation properly.

The cost that is not on the invoice

Your people's time, and it is front loaded.

Somebody in your business has to sit with us and explain how things really work, get us access to the systems, test the thing before it goes live, and run it for the first few weeks while it settles. That is real hours out of a small team that does not have spare ones.

We are not going to pretend that cost is zero, because the businesses where it gets treated as zero are the ones where the deployment quietly fails. But it is finite, it is heaviest at the start, and it is the mechanism by which the rest of it works. The alternative is a system built from a description of your operation rather than your operation, and that system will be wrong in ways nobody notices until it has been wrong for a month.

The question to actually ask

Not "can we afford this." Ask "what is this already costing us, and do we know or are we guessing."

If you know, the arithmetic will tell you the answer and you will not need a salesperson to interpret it. If you are guessing, then the first thing worth buying is not a build. It is a month of counting.

What to do with this

1. Do the two deal test. Fee divided by profit per job. Can you name that many losses last month?
2. If the answer is "I don't know," spend thirty days counting instead of thirty thousand dollars building.
3. Finish the sentence. Every week someone here spends hours doing ______. No sentence, no scope.
4. Name the owner before you name the budget. If nobody has time to own it, nothing else matters.
5. Below the line, spend a hundred dollars and an afternoon. The way it half works is your specification.