# What to charge for AI work without lowballing yourself > Source: > Published: 2026-06-19 00:00:00+00:00 The Ledger · Pricing # What to charge for AI work without lowballing yourself Hourly, fixed, or value based: how to price AI builds and automations so that being fast does not turn into being cheap. The fastest way to lose money with AI is to price the way you priced before AI. You used to spend two days on a thing, so you charged for two days. Now the same thing takes you two hours, so you charge for two hours, and you just gave away most of your income to be more productive. That is backwards. Here is how to set a number you can actually defend. ## Why hourly punishes you for being fast Hourly billing ties your pay to how long something takes. AI cuts how long things take. So the better you get with these tools, the less you earn per project. That is a trap, and clients feel it too: nobody wants a bill that goes up when you work slower. Hourly is fine for open ended work where the scope genuinely is not known yet, a research spike, a “help me figure out what is even possible” session. For anything with a clear finish line, price the finish line, not the clock. ## Three ways to price, and when each one fits **Fixed price per outcome.** You agree on a deliverable and a number up front. Good when the scope is clear: “an automation that turns these emails into CRM entries,” “a chatbot trained on your docs that answers the top 20 support questions.” The client knows the cost, you keep the upside of working fast. This is the default for most AI build work. **Value based.** You price against what the result is worth to the client, not what it costs you to make. If an automation saves a team 15 hours a week, that is worth far more than the afternoon it took you to wire up. Value pricing needs a number you can point to (hours saved, leads handled, tickets deflected) and the nerve to ask. It is where the real money is, and where most people undercharge by an order of magnitude. **Retainer.** A flat monthly fee to keep things running, tune prompts, handle the model changing under you, add the next small thing. AI work is not “ship it and walk away,” models shift and break, so ongoing care is a real service worth charging for. Retainers turn one project into steady income. ## How to set a number you can defend Start from the client’s side, not yours. Ask what the problem currently costs them: the salary hours going into the manual version, the leads they drop, the customers who churn waiting on slow replies. That number is your ceiling, and it is usually higher than the price you were about to name. Then sanity check from your side. Whatever a fixed price comes out to, make sure it clears what you would want per hour of your actual time, including the unglamorous parts: scoping, revisions, the model breaking the week after launch. If it does not, the scope is too loose or the price is too low. A few working rules: - Name the price as one number for the outcome, not a rate times an estimate. “This is $X” lands better than “it is $Y an hour and I think about Z hours.” - Quote a range before you quote a number, and let the client’s reaction tell you where you stand. - Build revisions into the price (two rounds, say), and put a line on what counts as new scope. Scope creep is where fixed price jobs go to die. - Charge a deposit. Half up front is normal, and it filters out the people who were never going to pay. ## When to walk Some clients want the AI discount: they read that the tool is cheap, so they think the work should be too. They are not your clients. The value is in knowing which tool, wiring it up so it does not embarrass them, and being there when it breaks. If someone only wants to pay for the API bill, let them go build it themselves. The real version of pricing is not a magic number. It is refusing to let a tool that made you faster make you poorer. Price the outcome, show the value, and keep the upside of being good at this. No course-shilling, no "DM me to learn how." If a number can't be shown, we don't print it.