Fixed Price vs. Hourly: How AI Agencies Actually Quote Projects

Two agencies can look at the same project and hand you two totally different kinds of quote. One says “$120 an hour, we'll bill as we go.” The other says “$8,000, delivered in four weeks.” Neither is automatically the better deal — they move the risk to different places, and it helps to know where before you sign.

Hourly: flexible, but the risk is yours

Hourly billing makes sense when nobody's totally sure yet what the finished thing looks like — early exploration, research, a project that's genuinely open-ended. The tradeoff is that the final number is unknown until the work is done, and if the scope grows quietly along the way, so does the invoice. You're carrying the risk of the project running long.

Fixed price: predictable, but the scope has to be real

A fixed price only works if the project was scoped properly first — a clear task, clear systems, a clear definition of done. Done right, it means you know the exact number going in, and the agency carries the risk of underestimating the work. Done wrong — quoted without real scoping — it usually means the agency pads the number heavily to protect themselves, or worse, cuts corners once they realize they underquoted.

What to actually ask before choosing

• If it's fixed price, ask what happens when you request something outside the original scope — a good answer is “we quote that separately,” not silence or a vague “we'll figure it out.”
• If it's hourly, ask for a not-to-exceed estimate anyway. Most agencies can give you a range even without a formal fixed quote.
• Either way, ask what's included after launch. A price that only covers the build and stops the day it ships is a different deal than one that includes a support window.

Which one we'd pick

For a well-defined, single-purpose agent, fixed price is usually the better deal for the client — you know your number, and the agency has already done the work of thinking through the scope before quoting. Hourly makes more sense for genuinely exploratory work where nobody could honestly write a fixed scope yet.