Pricing AI work: setup fees, retainers and usage-based plans

The five hundred dollar trap. Picture this. You build an AI agent that answers messages and books appointments for a hair salon. The owner loves it, pays you five hundred dollars once, and you shake hands. Three months later the salon is busier, and the agent handles thousands of messages a month. You check its conversations, fix what breaks, and pay its AI bill, and nobody pays you for any of it. On top of that, the owner sends you a small change request every week. You're now spending your own hours and money to run a service you sold for a single fee. This happens to many new builders, and it's almost never a technical problem. It's a pricing problem, and today we'll fix it.

What you will learn. By the end of this lesson you'll be able to do three things. First, split any AI automation price into three parts: a setup fee, a monthly retainer, and a usage charge. Second, calculate a real monthly bill for a client from a few simple numbers, using a small spreadsheet. Third, check that price against what the automation earns or saves for the client, so you can defend it with confidence. You don't need any finance background for this. If you can multiply two numbers, you can price AI work properly.

Three jars of money. Think of your pricing as three jars on a shop counter. The first jar is the setup fee. It fills once, at the start, and pays for designing and building the automation. The second jar is the retainer. It gets a fresh deposit every month and pays for keeping the agent healthy, watching it, fixing it and improving it. The third jar is usage. It fills drop by drop, with every message the agent handles, so it follows how busy the client really is. Most good AI offers use all three jars together.

The setup fee. The setup fee pays for the work before the agent goes live. That's the discovery call, mapping how the business answers messages today, building the workflow, and testing it with real examples. Most of the effort is connecting things: their inbox, their calendar, and their CRM. CRM means customer relationship manager, the tool where a business keeps its list of customers and leads, like a digital address book with notes. To price setup, estimate your hours and multiply by your hourly rate. Then add a buffer, because client systems always hide a surprise or two, like a missing login or a messy calendar.

The monthly retainer. A retainer is a fixed monthly fee for ongoing care, a bit like a gym membership or a maintenance contract for a boiler. Every month you watch the agent's conversations, fix anything that broke, tune the prompts when the salon adds a new service or changes its opening hours, and send a short report. Write the scope down clearly. For example, the retainer includes two small changes a month, and bigger changes are quoted separately. Without that limit, the weekly change requests from our opening story quietly eat all your profit.

Usage-based charges. AI models charge by the token. A token is a small piece of a word, roughly three quarters of a word on average, and you pay for every token the model reads and writes. It works like a taxi meter: the longer the ride, the bigger the fare. So when the salon gets more messages, your AI cost goes up. Usage pricing passes that cost on fairly. You charge per message or per conversation, at a price above your real cost. The markup covers your risk, the monitoring, and the months when costs spike.

Worked example: the salon agent. Let's price the salon properly. We estimate twenty hours of build at sixty dollars an hour, so the setup fee is one thousand two hundred dollars, paid once. Next, usage. The salon gets about three thousand messages a month, and we charge one cent per message, so the usage charge is thirty dollars. Now add the retainer of three hundred dollars, and the monthly bill comes to three hundred and thirty dollars. Finally, check your real AI cost: those messages at under half a cent each come to just twelve dollars, so usage is safely covered.

Check the price against the return. So is three hundred and thirty dollars a fair price? Look at ROI, short for return on investment: what the client gets back for what they pay. The agent costs them three hundred and thirty dollars a month. It saves about forty hours of front desk time, which at fifteen dollars an hour is six hundred dollars. It also answers late at night, so around nine missed bookings recovered, at fifty dollars each, add four hundred and fifty more. That's about three times what they pay. When you can show that, price objections mostly disappear.

Package it into tiers. Clients find it hard to pick from a formula, so wrap your three jars into simple tiers. Starter covers up to one thousand messages with a lighter retainer, just one small change a month, for two hundred and fifty dollars a month. Growth covers three thousand messages for three hundred and thirty, which is exactly our salon. Pro covers ten thousand messages for seven hundred, with faster support. Each tier includes a retainer sized to it and a message cap. If a client goes over the cap, they pay an overage, meaning one cent per extra message. Most people choose the middle option, so make Growth your best fit for a typical small business.

Common mistake: unlimited for one price. The most common mistake is selling unlimited use for one flat price, either once or monthly. It feels generous and easy to sell. But your AI meter keeps running, and every new customer message costs you money, while your income stays a single coin on the desk. A busy online shop running a promotion can multiply messages overnight. The fix is simple and fair: put a message cap in every plan, an overage price above the cap, and the retainer scope in a written agreement before you start the build.

Recap. Let's recap. First, the setup fee pays for discovery, building and testing, and you agree it once, in the contract. Second, the retainer pays for ongoing care every month, marked on the calendar, with a clear limit on changes. Third, usage follows the meter: price each message above your real AI cost, and protect yourself with caps and overage. Put those together, check them against the client's return, and you have a price that's fair to the client and keeps your business healthy as they grow.

Your exercise. Now it's your turn. Pick one process from a small online business you know, for example qualifying leads for an online course seller, or sending invoices for a cleaning company. Estimate your build hours, the messages or tasks per month, and the AI cost per task. Then build the same small sheet we used: setup fee, usage charge, retainer, monthly bill, and your real AI cost. Finish by writing one sentence about the client's return. In the next lesson, we'll turn those numbers into a dashboard your client can actually see.