If you do contract work, AI has probably already changed your economics without you renegotiating anything. The first draft that took four hours takes forty minutes. The boilerplate module writes itself. The research pass that filled a whole afternoon is now a long coffee break.
That’s great news for your hourly life and terrible news for your hourly invoice. Bill by the hour and every efficiency gain hands the saving straight to your client. Meanwhile the parts AI can’t do — judgment, taste, accountability for the result — are exactly the parts you were never billing separately for.
The fix isn’t to hide the AI or to pad the timesheet. It’s to bill for the outcome and be straightforward about how it gets made.
When invoicing for AI-assisted work, stop billing by the hour and bill for outcomes: fixed project fees, deliverables or milestones. Price the review and accountability, not the drafting, because checking the output is where the time and risk now sit. Settle disclosure, confidentiality and ownership in the contract up front, and keep invoice line items about deliverables, not process.
Stop billing the hours AI gave back
The clearest signal that hourly billing has stopped working is when your best work produces your smallest invoice. A fixed project fee — a single agreed price for a defined result, however many hours it takes — priced from the value of the finished thing, removes that perverse incentive entirely. The client gets a number they can approve up front; you keep the upside of being fast. If you’ve never priced that way, how to set freelance rates is the place to start.
If you can’t move fully off hourly — some clients genuinely can’t buy any other way — move the unit up instead. Bill per deliverable, per sprint, per milestone, or per week of engagement. Anything that decouples your fee from the stopwatch will do.
Price the review, not the generation
Here’s the useful mental model for AI-assisted contract work: generation is cheap, verification is not. Producing a plausible draft is now nearly free. Deciding whether it’s right — checking the facts, catching the subtly wrong logic, making it fit the client’s context and standards — is where all the real time and all the real risk now live.
So when you scope a project, estimate the review effort honestly rather than the drafting effort. A deliverable that AI can draft in minutes might still need hours of checking, and that checking is the thing your client is actually paying you for. Quotes that assume “AI does it, so it’s quick” are how people end up doing unpaid QA on their own work.
What your contract should say about AI
AI-assisted work adds a few clauses that are worth settling before the project starts, not after a dispute. They sit alongside the basics every contract for a remote client should already cover:
- Disclosure and permission. Some clients are fine with AI tools, some have policies against them, and a few are contractually barred by their clients. Ask before the project starts. A one-line acknowledgement in the agreement is enough, and it protects you completely.
- Confidentiality. Be explicit about whether client material goes into a third-party tool, and which one. If you’re under an NDA, check that using a hosted model doesn’t breach it — and keep sensitive client data out of tools you don’t control.
- Ownership and warranties. State who owns the delivered work and what you’re actually warranting. You can reasonably warrant that you reviewed the output and that it’s fit for the agreed purpose. Be careful about blanket originality guarantees for anything a model generated.
- Accountability. Whatever produced the draft, you’re accountable for what you deliver. That’s the value you’re charging for, and it’s worth saying out loud in the proposal — it’s a selling point, not an admission.
What to put on the invoice
Your invoice should describe outcomes, not process. “Landing page copy — three variants, two revision rounds” is a line item a client can approve without a conversation. “12 hours, AI-assisted drafting” invites a negotiation you don’t want and shouldn’t need to have.
A few practical habits:
- Bill in deliverables the client recognises. Match the line items to what you promised in the scope, so the invoice reads as proof of delivery.
- Invoice milestones as they land, rather than saving everything for the end. Faster delivery should mean faster payment, not one big invoice at the finish line — otherwise AI speeds up your work and nothing else.
- Keep the revision boundary on the document. If the scope said two rounds, note it. Extra rounds become an easy, unawkward add-on line rather than a favour — the simplest defence against scope creep.
- Don’t itemise your tool costs. Your subscriptions are a business expense, not a client pass-through, and listing them just draws attention to the wrong thing. They usually belong with your other tax-deductible expenses instead.
Reprice the work that got faster
If you’re on a long-running contract that was priced before you changed how you work, revisit it. Not to charge more for less — to move to a footing that survives the next efficiency gain. A retainer that was scoped as “20 hours a month” makes no sense once the same output takes eight; convert it to a scope of deliverables at a fee both sides think is fair, and the hours stop being the subject of the conversation. Retainers vs project fees walks through which structure suits which kind of work.
The clients worth keeping care about the result, the reliability and the fact that a named person stands behind the work. None of that got cheaper.
Frequently asked questions
Do I have to tell clients I use AI tools?
You should tell clients you use AI tools before the project starts, even where no law requires it. Some clients have policies against them or are bound by their own clients’ contracts, and an NDA may restrict putting their material into a hosted tool. A one-line acknowledgement in the agreement settles it and protects you if questions come up later.
Should I charge less because AI makes me faster?
No — charge for the value and reliability of the result, not the time it took to produce. The client is paying for a correct, reviewed deliverable and for someone accountable for it, and none of that got cheaper. Passing every efficiency gain on as a lower fee just trains clients to expect your price to fall each time your tools improve.
Who owns work created with AI tools?
Ownership between you and the client is whatever your contract says, so state who owns the work and when it transfers. The harder question is copyright: in some countries, including the US, purely machine-generated material may not be protected at all, and the law is still developing. That’s why it’s wise to warrant that you reviewed the work rather than guarantee its originality.
How do I estimate a fixed fee for AI-assisted work?
Estimate the review, testing and revision effort honestly, then price from the value of the finished deliverable rather than the drafting time. Use past projects to see where checking and client feedback actually took the hours. Add an explicit revision limit, so extra rounds become a paid add-on rather than an open-ended commitment.
Where Billable comes in
Outcome-based, milestone-billed contract work only pays off if the invoicing keeps up with the delivery — and that’s what Billable is for. Send a clean, deliverable-based invoice the moment a milestone lands, track what’s outstanding across every project, and let overdue payments get chased automatically instead of sitting in your head.
It’s built for a business of one, at a flat $9 a month. Get paid as fast as you deliver.