How to Quote a Project When AI Does Most of the Work

Here is the arithmetic that makes this urgent. You used to bill a build at 40 hours and 100 per hour, so 4,000. With AI the same build takes 12 hours. Quote it hourly and you are asking for 1,200 for identical delivered work. You have become three times more productive and taken a seventy percent pay cut for it. Learning how to quote a project when AI does most of the work is mostly about noticing that hourly billing converts your efficiency gain directly into a client discount, and then stopping.

The hourly trap, in numbers

The third column is not greed. The client's outcome is identical across all three, which means the value they receive is identical, and the only thing that changed is your cost of production. In every other trade, a supplier who finds a cheaper process keeps some of the margin. Hourly billing is unusual in handing all of it to the buyer automatically.

There is also a second-order problem. If you quote 1,200, you have set the market price for that deliverable at 1,200 for that client permanently, and they will anchor there for the next project and tell their network. Discounting once is a pricing decision you make for years.

What the client is actually buying

Four things, and AI compresses one of them.

1. The working thing. This is the part AI makes faster. It is also the part with the least defensible value, because the client increasingly knows it is faster.

2. Knowing what to build. Turning a vague business problem into a specification that survives contact with reality. AI does not do this for someone who cannot already do it.

3. The judgement that it is correct. Knowing that the auth flow has a hole, that this will fall over at 10,000 rows, that the data model will not survive the second feature request. This is the expensive part and it is entirely yours.

4. The guarantee. Someone accountable when it breaks at 2am. AI cannot be on the hook, and being on the hook is worth real money.

Quote the four things, not the hours. The build got faster; the specification, the judgement, and the accountability did not, and those are three quarters of what the client needs even if they only asked about the first one.

How to quote a project when AI does most of the work

Price the outcome as a fixed fee, and present it in terms of the deliverable rather than the process. A structure that holds up:

• A fixed price for a defined scope, with the scope written as observable outcomes rather than tasks or hours.

• An explicit list of what is included: the build, the specification work, testing, one round of revisions, handover.

• A named support period afterwards, priced separately so it reads as a real thing rather than a freebie you will resent.

• A change mechanism. Scope changes are quoted as additions, not absorbed, and saying so upfront prevents the argument later.

Do not itemise hours anywhere. The moment hours appear in a quote, the conversation becomes about hours, and you cannot win that conversation once the client knows a machine is doing part of the typing. If you have not set a base rate for this kind of work at all yet, pricing an AI product from first principles is a reasonable starting point before you get to project-level quoting, and it sits inside the wider question of how to monetise AI work.

When they ask whether you used AI

They will, increasingly. The answer is yes, stated without apology, and immediately reframed.

Something close to: yes, I use AI for the parts it is good at, which is why this is four weeks instead of three months. What you are paying for is knowing what to build and being certain it is right, and that part has not got any faster. Then move on. The defensive version of this answer invites a discount request; the matter-of-fact version usually ends the topic.

Whether to volunteer it unprompted is a separate judgement call, and the arguments run both ways depending on the client and the contract. Decide it before it comes up rather than during a pricing call, and if you take on AI work regularly, a written AI usage policy settles it once instead of every time. The same upfront-honesty logic applies when a client's idea will not work, which is a conversation worth having early for the same reason.

A worked example

A client wants an internal tool: staff submit expense claims, a manager approves them, approved claims export to the accounting system. Previously a six week build. With AI, realistically two weeks of your time.

The wrong quote itemises two weeks at your day rate and lands around 5,000. The right quote prices the four things separately in your own head, then presents one number:

Three of four columns are unchanged. The quote goes out as a single fixed fee for the outcome, with the inclusions listed and the support window named, and nothing in it mentions days. If the client pushes on price, the negotiation is about scope, which claims types are in or out, whether the accounting export is included, not about how long you spent typing.

The discipline is to decide the number before you know how long it will take you. If the price is right for the outcome, it stays right whether the build takes two weeks or four days.

The one case for charging less

Honesty requires this section. There is a real scenario where your price should fall: when the client could now plausibly do it themselves with the same tools and the job genuinely is just the build. Simple marketing sites, basic forms, straightforward content work. The value of the specification and judgement layers is small there, because the problem is not hard to specify and not easy to get wrong.

For that work the market price will fall and no quoting technique will stop it. The response is to move up the stack rather than to defend the price, because the defensible margin is in the problems where getting it wrong is expensive. Those are not getting easier and the tools have not touched them.

FAQ

Should I charge less because I used AI?

Not for the same delivered outcome. The client's value is unchanged, and hourly billing is the only common arrangement that passes a supplier's efficiency gain straight to the buyer. Price the outcome and keep the margin, as any other trade would.

Do I have to tell the client I used AI?

Check your contract first, since some client agreements now require disclosure or restrict it. Absent a contractual term it is a judgement call, but answer honestly when asked, because being caught being evasive costs far more than the disclosure would have.

How do I price a fixed fee when I am not sure of the scope?

Sell the scoping as a small paid piece of work first, delivering a written specification and a fixed quote for the build. It protects you from underquoting an unclear project and it demonstrates the part of your value that is not the typing.

What if a competitor quotes a third of my price?

Some will, and some clients will take it. Compete on the parts AI does not do: a specification that holds, judgement about what will break, and accountability afterwards. If a client genuinely only needs the build, that work is commoditising and chasing it down is not a strategy.