The Numbers · AI Scaling

The numbers for AI Scaling · the working math

The numbers behind the business,
shown in full.

The business: an agency that installs AI agents for local companies. We build it for you, you run it. The Case argued why it exists. This document breaks down the math, one client at a time, until the whole year makes sense. Every assumption printed, and where the model touches reality, reality is drawn in.

Model figures approved by AI Scaling, printed conservative · zero compounding assumed · your results depend on your execution

The math begins
01 One Client The atom of the business: both sides of the trade.

Start with one client. Everything else is multiplication.

Your client, a local business buying an AI agent from your agency, pays $8,000–$15,000 to install their system, then about $2,250 a month, roughly $38,500 across their first year, priced to what their industry already pays for the problem, not to hope. Two costs come out of it: the cost of doing the client work, which the platform automates, and the 10% platform fee.

A The anatomy of one client Model figures · first year
Contract · year one $27,000 $2,250 a month, twelve months
Delivery (8%) −$2,160 your cost to run it: compute, hosting, usage
Platform fee (10%) −$2,700 our share, the only cut we take
You keep $22,140 82% of the contract
Why the cost of the work is 8%, not 40%: a traditional agency pays people to do the client work. Here the platform does that work. That ninety-percent-range margin on the work is a measured figure from our own operating history, not a projection. Printed conservative: this breakdown shows the monthly payments only. The $8–15K install fee lands on top of it.

And the client? They're not doing you a favor at $2,250 a month. Here's the trade from their side of the table:

B The client's math Why the contract renews itself
Client pays $2,250/mo patient-intake agent, run by you
Agent recovers ≈$24,000/mo ~22 missed appointments a month, booked instead of lost
Client's return ≈10× every month the agent runs
Reading note: representative example: a dental practice whose front desk missed roughly a third of inbound calls at ~$1,100 per new patient (the return rounds down to 10×). Different niches, different math, same shape: the agent pays for itself first. Walked through live on your call.

You're not selling a service that costs your client money. You're selling one that makes it. That's why the contracts hold.

02 The Cost What a client costs to get.

One client pays for the next seventeen.

The arithmetic is short: a qualified sales call costs about $250 in ad spend to book, and at a 20% close rate it takes five calls to sign one client. Five × $250 = $1,250 to sign a client who leaves $22,140 in your pocket. What that one client leaves you pays to find the next seventeen, with change to spare.

Timing works in your favor too: you pay the $1,250 up front, but what you keep from that client comes back at about $1,845 a month, so a single client pays back what it cost to find inside the first month, then keeps paying.

And the meter doesn't start at zero. The first $1,000 of marketing is included (your first four booked calls are on us), and your first client is guaranteed inside 90 days, in writing. Your start is promised in the contract.

C One client in, seventeen funded 17.7× on the cost to sign one
Reading note: $250 per booked call and a 20% close rate are the operating model we coach you to, and we've printed both on the cautious side. Booking the calls is what the customer-finding system we build for you is for; closing them is what the twice-daily live training is for. Pressure-test both on your call. We put the operating data behind them on screen.

When one client funds the next seventeen, the question stops being "does the math work" and becomes "how many can I sign."

03 The Year Three paces vs. the floor, with reality drawn in.

What year one looks like, month by month.

Monthly payments start small and stack up. You spend year one signing clients, and each one pays as they go, so the cash you collect in twelve months trails the business you're actually building. Here is that cash, at three paces, against the floor the guarantee holds. The curves bend upward because the monthly payments stack: every client you sign keeps paying while you add the next.

At these figures, $3,000 a month of marketing signs about two clients a month: the base case, and the pace one partner runs solo. It collects $627,000 in year-one cash: install fees up front, then the stacked monthly payments doing the rest. The first client is guaranteed inside 90 days, so even the slowest start has a start date. And notice what the install fees do to the slow line: even one client a month clears the floor. The guarantee exists for the year that stalls harder than that. Chapter 06 prints it.

D Year-one cash · three paces vs. the floor $250,000 in writing
Year-one cash 3 clients / mo · $941K 2 clients / mo · $627K 1 client / mo · $272K ~0.9 / mo · lands exactly on the floor
For scale · first-year cash: the model vs. what's been recorded
Floor crossings (the dots on the floor line): three clients a month clears in month 5; two, in month 7; even one a month clears in month 12. The dotted curve is the break-even pace: about 0.9 clients a month collects exactly $250,000 in year one. Sign faster and you beat the floor on cash; stall below that, and the guarantee tops up. The strip above places the model against reality: DeepOps ($1.6M, recorded; our own operation) and Doriane (~$1M, reported by her) both ran ahead of the base line.

And year-one cash is the smallest number this business produces. At two clients a month you finish the year holding twenty-four paying clients: $648,000 a year arriving every month walking into year two. That's the honest trade of a build year: it pays you $627,000 while you assemble a machine that collects about $1.16 million the year after. Chapter 05 shows that math.

The guarantee isn't the ambition. It's the floor under your build year, and the year after pays on everything you built in this one.

04 The Keep What's left after honesty.

What you keep, after every real cost.

Let the client list fill up. A full year at the base case, twenty-four clients paying monthly, the $648,000 a year you exit year one collecting, charged for everything: the automated client work, the platform fee, a full year of marketing, and the tools that run the shop. Most margin math in this industry forgets half of these. This page charges all of them.

E A full year at the base case Steady state · all costs
Reading note: that's a 75% all-in owner margin, pre-tax, after the client work, after our fee, after your marketing, after your tools ($6,000/yr for CRM, customer-finding tools, and agent hosting). Marketing is charged in full even though the client list is held flat here: its new signings counted at zero, like everything in chapter 05. The margin on the client work alone runs in the ninety-percent range because the machine does the work.
Your hours 10–15 / wk sales conversations and decisions. Delivery is the machine's job. Live training runs morning and evening, so it fits around a life.
Your ceiling ≈25–30 clients what one partner runs before a first hire. The base plan peaks at 24 concurrent, inside it; the fast line crosses it around month nine. Growing past it is a hiring decision, not a rescue.

Delivery is automated. That is the entire secret of the margin.

05 The Excluded Upside Everything the model refuses to count.

What we left out of the math, on purpose.

Every chart above assumes the business never compounds. That assumption is false on purpose. It keeps the math conservative. Here is what was deliberately left out of every figure on this page:

F Left out of every chart above Counted at zero
Renewals
82% of clients renew into a second year. Keep the book you built, hold the same pace, and year two collects about $1.16 million in cash: $531,000 from the clients you already signed, plus the same build ramp again. Nearly double year one, from the same arithmetic. The year-one charts count it at zero.
Referrals
Every client whose operations stopped leaking money knows ten more owners like them. Partners tell us referrals become their cheapest way to find customers. Counted at zero.
Expansion
Clients start with one agent and add more as trust builds: intake, then scheduling, then reactivation. Contract values grow with scope. Modeled flat.
Reading note: renewal figure from our operating model; year-two cash ≈ 24 × 82% × $27,000 retained ($531,360) plus the same-pace ramp ($627,000): ~$1.16M. If you want the optimistic version of this document, add this box back in. We won't.

The fantasy isn't the top line of the chart. The fantasy is this box, and we left it out of the math.

06 · The bad year

Now assume it goes slowly. The floor still pays.

With install fees counted, even one client a month clears the floor. So assume worse: one client every other month: six clients all year, every one at the low end of the install fee. You collect $135,750 in cash. That's under the floor, and it's the one case the written guarantee is built for: we make up the difference to $250,000.

Exhibit G · a slow year Downside, printed
Base case
$627,000
A stalled year
$135,750+$114,250 covered

A stalled year lands under $250,000, and the striped span is the guarantee paying the difference, up to the $250K tick. That's the whole job of the floor.

Model figures reviewed & approved · AI Scaling · 2026
Your model · rebuilt with your niche, on the call
Stalled-year math: ~1 client every other month at the $8,000 install fee plus the monthly payment, cash collected in twelve months · the guarantee makes up any shortfall to $250,000 · who qualifies, how it's measured, and how we pay live in the agreement.

You don't need the good year to be safe. That's the entire point of a floor.

07 The Receipts Models promise. These were collected.

The math above is a model. These numbers were collected.

$32M+ collected by the people running businesses we've built
$1.6M DeepOps, year one · our own money, on this platform, first
$1.8M Alex Miranda (Virtrify) · current revenue, $3M+ lifetime
$3M+ Doriane Padilla (Fantasy Lab) · year-two revenue, $4M+ lifetime

Case-study results reported by the partners themselves: individual examples, not typical results or guarantees.

08 The Next Step Bring your questions.

Bring your questions to the call. We'll walk the math with your niche in it.

Thirty minutes: this arithmetic rebuilt around your background and your market, our operating data behind every figure, and the full terms, including the floor, in writing before anything starts. Every number this page doesn't print, including what the license costs, goes on the table there.

Who qualifies for the guarantee, how it's measured, and how we pay are defined in the licensing agreement.