How to Price AI Automation Services (Without Guessing)

New AI agencies usually price by gut: "What would feel reasonable?" The first number sticks. Early clients tell their network what you charged, your case studies carry that price, and raising it later means fighting your own reputation. Getting the first price right is worth an afternoon.

Start with a menu, not a quote

When a prospect asks "What would this cost?" and you answer "It depends," you've asked them to design the project for you. Most won't. A menu with three fixed offers does that work in advance:

Treat these as planning ranges. Where you land inside them comes from the two tests below.

Test 1: Anchor to the outcome

Write one sentence per offer: "This Build saves the buyer about $X over a year" or "produces about Y leads a month." Then price the offer at roughly 10 to 20% of that value. Newer agencies price near 10%; specialists with several case studies price near 20%.

Use the buyer's own numbers, not your optimistic ones. On the discovery call, ask two questions:

1. "Conservatively, how many hours does this take today?"
2. "What's your blended internal hourly cost for that work, including benefits?"

Worked example: a buyer says the task takes 14 hours a week at $45 an hour. That's about $32,760 a year. At 10 to 20%, the Build prices between roughly $3,300 and $6,500. The buyer can see exactly where the number came from because it's their math.

Test 2: Hold a profit floor

New agencies underestimate delivery time, often by 40 to 60% on early projects. So ask: "If this takes 1.5 times the hours I expect, do I still earn my minimum hourly rate?" A common floor in the US is around $75 per delivered hour, enough to cover taxes, tools and admin. Use your own local equivalent.

If the price fails, raise it (the outcome anchor usually supports it) or cut the scope until it passes. Never ship a failing price and hope to "make it up on volume."

Don't negotiate against your own menu

The menu only works if the prices are the floor. When someone asks for "a small custom thing for $750," either decline, quote custom work at double the menu rate, or route them into the Audit. A menu that bends on every call is just custom quoting with decorative prices.

Free audits usually backfire

A free audit trains the buyer to value your thinking at zero. A paid Audit at even $750 filters for serious buyers and naturally leads into the Build it recommends.