The HVAC owner is at the kitchen table with the P&L and a spouse who was not on the Zoom. Five thousand a month. Sixty thousand a year. The CPA will ask the same question tomorrow in a different tone. If he cannot say what the retainer costs, what it replaces, and what "done" looks like, he should not buy it. They will smell it. They should.
Applied Margin's embedded partnership starts at $4,000 a month. That is not the opening offer. The opening paid engagement is a Margin Scan, so there is a map before there is a monthly commitment. This is the plain-English version to use at the kitchen table when someone asks whether an AI consultant is worth it for a small business.
You're not the only adult in the money conversation. There is a spouse. There is a CPA. There is a partner who was not on the call. The sentence has to survive them.
Say the cost like a line item, not a dream
"We're considering $4,000+ a month for someone to own the next workflow in the shop. Not software. A person. We would not start there. We'd pay $1,500 for a 45–60 minute owner interview and seven-business-day analysis of our shop first. If that map is thin, we walk. If it's real, the monthly fee is the implementation, one job at a time."
That's the whole pitch. No "transformation." No "keeping up." Your CPA does not need a model name. Your spouse does not need a thread. Compare it to things you already understand: a part-time office hire you cannot find, plus the two months you'll spend training them on tribal knowledge. The agency retainer you already pay to keep ads on. The fractional CFO or fractional HR person you already decided was cheaper than a full-time seat.
If $4,000+ a month would bounce payroll, you are not the buyer. That's fine. The free 15-minute Mini Margin Check exists so you can learn the money / time / customer test and do it yourself. I would rather you buy the cheap seat than finance a retainer you resent.
What it replaces (and what it doesn't)
It does not replace the hygienist, the lead tech, the paralegal, or the store manager. Hands work stays. It replaces a slice of the office loop that currently eats owner time and still leaks money: missed-call follow-up, recall texts, first-draft estimates, intake notes, review replies, the Friday status email.
Efficiency is hours back. Effectiveness is money that was dying in the gap. Quality is the customer not getting a different experience depending on who was working. Those are the only three doors. If you cannot name which one this month is buying, the retainer is a hobby with an invoice.
Hypothetical (not a client, not a promise): you miss 8 inbound HVAC calls a week after 4 p.m. You close maybe 1 in 3 of the ones you do catch. Average ticket $3,800. That's not "AI will make you $8,000 a week." That's "if we captured even 2 extra booked diagnostics a month that turn into one job, the retainer is paid and then some." Write the assumptions on the page. If you don't miss calls, this math is zero. Then we look at estimate lag instead. I will not invent a shop I transformed. If a number isn't on your P&L yet, it is a model. Models have assumptions. Assumptions can be wrong.
Dental recall, law intake, agency status emails, and multi-location out-of-stock sheets use the same shape. Name the loop. Baseline it. Convert hours to dollars only if you will redeploy them onto billed work, stop hiring the next coordinator, or sell capacity you currently waste. If your books are already full eight weeks out, don't buy a retainer to fill a schedule you cannot seat. If one extra retained matter a quarter doesn't cover the year at a conservative close rate, don't force the story.
Risk, in language a CPA likes
The risk is not that the model gets smarter. The risk is that you pay for a science project. So sequence the commitment. First: a free 15-minute Mini Margin Check to decide whether a deeper look is warranted. Second: the $1,500 Margin Scan, with a 45–60 minute owner interview and seven-business-day analysis. You get a ranked map in your language, with every assumption listed. Third: only then, an embedded partnership from $4,000 a month. One active workstream at a time. You can see what shipped. The map remains useful even if the monthly relationship never begins.
What you should not do: sign a year because a vendor demo was slick. What I will not do: a free audit, a free "discovery," or a guarantee tied to hours I haven't measured. Risk sits on finding a real loop in the assessment, not on a movie-trailer transformation.
Your CPA will ask about cash vs. capitalized software. This is a service. It's an operating expense, like the bookkeeper. There is no custom platform to depreciate unless we later decide a rare build is justified — and for most shops it isn't. Your spouse will ask "what if it doesn't work?" The honest answer: we kill a loop that doesn't move Effectiveness, Efficiency, or Quality, and we pick the next one. We do not keep paying for a helper nobody uses.
What "done" looks like
Done is not "we are an AI company." Done for month one is: one hated loop is mapped, shortened, and running with a human in the send seat. The office manager can run it on Tuesday without me in the chair. You have a before/after you can show: time, money, or fewer customer surprises.
Done for a quarter is: two or three loops, a folder of shop context so the helper doesn't sound generic, and a list of what staff said broke. At renewal you should be able to scroll what shipped. If you can't, don't renew. Done is not a dashboard with login counts. Seats are procurement. A working recall text, a working first-draft estimate, a working intake note — that's the artifact. The 7-minute video from the assessment is how you get to "done" without translating a 12-page PDF at dinner. Watch it together. Pause on the assumptions. If the spouse or CPA says "this assumption is garbage," they're right to stop you. Better now than on month four of a retainer.
Use this. Change the numbers to yours.
Kitchen-table script (say it out loud):
I don't want another $20 tool we'll ignore. I want someone who will sit with how a job actually moves and install one workflow at a time.
First we pay $1,500 for the Margin Scan. Here is the ranked map and the assumptions behind it.
The monthly fee would be $4,000+ ($48,000+/year). I'm comparing it to [missed calls / unbooked recare / unbilled admin hours], using conservative math, labeled as hypothetical until we baseline.
If the first loop doesn't move, we don't invent a second science project. We stop or we change the loop.
Human still sends every customer-facing message. No one is getting fired. The trucks, the chairs, the matters, the registers still need people. This is the office mess.
If you cannot say that without wincing, you are not ready. Start with the Mini Margin Check and come back when the sentence is clean.
Questions I would want your CPA to ask me: What is the first loop, in one sentence? What is the baseline we will write down in week one? What will we not automate (patient email, court filings, anything that can harm a person if it goes out wrong)? When do we review kill vs. keep? Is this replacing a hire, delaying a hire, or just buying back owner nights? What happens if we cancel — do we keep the files, the prompts, the map? If I cannot answer those, don't hire me. If your current "AI guy" cannot answer those, don't hire them either.
Effectiveness, Efficiency, Quality. That's the whole ROI conversation. More booked work, hours back that you will actually use, a customer who gets the same shop on a Tuesday that they got on a Thursday. If a retainer cannot name which of the three it is buying this month, it is a hobby with an invoice.
FAQ
Is an AI consultant worth it for a small business? Only if you can say the monthly number, what loop it replaces, and what done looks like on Tuesday. If you can't, buy a free Mini Margin Check or a paid diagnostic — not a retainer.
How do I explain $4,000+ a month to my CPA? It is an operating expense for owning an active improvement workstream. Sequence the commitment: free Mini Margin Check, then the $1,500 map, then monthly implementation only when the map supports it. No custom platform unless a build is later justified.
What if it doesn't work? Kill the loop that didn't move money, time, or customer experience. Don't invent a second science project. Keep the map and the files. Don't renew a hub that's empty.
The next step is the paid assessment: $1,500 for a 45–60 minute owner interview and seven-business-day analysis. You get the map and the seven-minute video you can forward. If you only want the filter, the free Mini Margin Check is the door. The $4,000+ month comes after you've already bought twice and seen your own numbers. Bring the spouse. Bring the CPA. I would.