Pricing AI-assisted work: what to charge when AI cuts the hours

Last year a set of annual accounts took your senior eleven hours. This year, with AI drafting the workpapers and the notes, it took six. If you bill by the hour, you just cut that client’s price by almost half, and nobody in the firm decided to.

That is the problem with hours-based pricing once AI is in the workflow: the price is tied to the one input AI is best at shrinking. This piece is for an owner who has seen the time drop and wants to decide, on purpose, what happens next.

An hourly fee is a formula: time multiplied by a rate, so every efficiency flows straight to the invoice. That was tolerable when efficiencies arrived slowly. AI moves faster: a review that took ninety minutes now takes twenty.

So the hourly firm either raises the rate to keep revenue flat, which looks odd to a client, or keeps it and watches revenue fall as the team gets better. Both are the formula deciding for you.

Plenty of firms have already moved. In Xero’s Modern Practice Playbook survey of 524 senior accountants and bookkeepers in Australia, released 29 September 2026, firms said they bill an average of 69% of client engagements on non-hourly terms. That survey was commissioned by Xero and drew partly on Xero’s own customer base, so read it as a picture of Xero practices, not the whole profession.

Learning check

What would you do? Pick one; every answer is explained.

Last year a client's annual accounts took your senior eleven hours. This year, with AI drafting the workpapers, it took six. You bill by the hour, so the invoice is roughly half what it was. What is the best way to read that?

1. a Your billing formula cut the price, and that needs a deliberate decision.
2. b A clear win: the client pays less and your rate holds.
3. c Quietly lift the hourly rate so this job's revenue holds.

A colleague quotes a Xero survey in which Australian practices said they bill an average of 69% of engagements on non-hourly terms (such as fixed fees). They call it proof that the whole profession has moved off hourly billing. How should you treat the figure?

1. a As a reliable figure for every firm in Australia.
2. b As a picture of Xero practices, not the whole profession.
3. c Ignore it, as a vendor survey tells you nothing.

Your decision for this step (saved with your progress, and in your summary)

Mark step 1 done

Moving to fixed fees does not solve this by itself. Most fixed fees were built from hours. Someone looked at last year’s time, multiplied by a rate, rounded, and called it a package. The hours were just hidden inside the number.

Once AI arrives, the next package built from fresh time data comes out lower, because the hours fell. And the scoping assumptions go stale: a package priced around “about eight hours of cleanup” no longer describes the work, so you cannot tell when a client drifts out of scope.

Our piece on AI drafting tax returns shows how to split your work into the mechanical part and the judgement and risk you carry. Do that first. The next question is what to build the price from instead.

Build the price from three things, in this order

1. The outcome. What does the client get, and what would it cost them not to have it? Lodged on time, no penalties, a clean loan application, a decision made with real numbers.
2. Your responsibility. Your name, registration and insurance sit behind the work. AI does not change that, and the risk does not shrink when the drafting gets faster.
3. Your cost floor. Time, software (including the AI tools themselves), review and overhead. This is the number below which you do not go. It is a floor, not the price.

If you only remember one thing: hours tell you whether a job is profitable. They should not tell you what it is worth.

Sometimes as accountants we feel empathy toward business owners, and we underprice them and then can’t increase those prices later.

Learning check

What would you do? Pick one; every answer is explained.

Each year you refresh your fixed fees from the past year's time records. This year the hours are much lower, because AI now drafts much of the work. What is the best move?

1. a Drop fixed fees and go back to billing by the hour.
2. b Hold fees while you rebuild packages around outcomes.
3. c Refresh them as usual, so each fee matches the work.

You priced a monthly bookkeeping package around 'about eight hours of cleanup'. With AI, the cleanup now takes far less time, but the client sends you much more to process than when you priced it. What does this show?

1. a The package is overpriced and should be cut to match.
2. b Nothing is wrong, as the work still fits in eight hours.
3. c Hours no longer describe the job, so you cannot see scope creep.

You have worked out the full cost of delivering one of your fee packages: staff time, review time, overheads and the AI subscriptions. How should you use that number?

1. a Add a margin on top and make that the price.
2. b Leave out the AI subscriptions so it shows staff time.
3. c As the floor you won't go below, not as the price itself.

Your decision for this step (saved with your progress, and in your summary)

Mark step 2 done

Holding your price while the time falls works best when the client gets something the hours used to crowd out. The saved time needs a destination, written down, before it disappears into more of the same work.

Xero’s survey asked Australian practices exactly this. Here is where they said AI-freed capacity is going.

Show the numbers

Item	Value
Higher-value advisory work	44%
Innovation and new workflows	40%
Training and upskilling	39%
Business development	36%

Source: Xero, Modern Practice Playbook (Australia), 29 September 2026

Xero’s UK edition, published 4 August 2026 from 520 UK practices surveyed in May, reported three in five practices directing AI time savings towards advisory. Those are intentions, not results. The firms that make them real do three things.

• Name the work“More advisory” is a wish. “A forty-minute planning call before 30 June for every client on the middle package” is a plan.
• Put it in the packageIf the client never sees the new work, they will only see the old work getting faster.
• Keep some of itCapacity you do not fill can mean fewer late nights in season, or taking on clients without hiring. That is a legitimate use, and in the same Australian survey 35% of practices said AI let them maintain or grow output without adding headcount.

Learning check

What would you do? Pick one; every answer is explained.

AI has freed up hours across your team, and everyone agrees the time will go to 'more advisory'. What is the best next step?

1. a Name the work, such as a planning call before 30 June, and add it to a package.
2. b Leave it open so each person uses the time as they see fit.
3. c Do the extra work, but keep it out of the package as a bonus.

AI has saved your team hours. You would rather use some of that time to cut late nights in tax season than fill it all with new work. Is that a sound choice?

1. a No, every saved hour should become billable advisory.
2. b Only if you also cut the client's fee to match.
3. c Yes, keeping some capacity is a legitimate use of it.

Your decision for this step (saved with your progress, and in your summary)

Mark step 3 done

The biggest pricing gain from AI for a small firm is often better scoping, because most underpricing happens at the first quote.

AI is good at the unglamorous part: turning a discovery call into a structured scope, flagging what the call did not cover, and checking a draft engagement against the scope you actually agreed. Anderson Peter George’s build shows how to set up a pricing project with your own framework, and our onboarding guide covers the steps after the client says yes. The rules that keep it safe are short.

• AI drafts the scope. A person who knows the client confirms it.
• AI can suggest where a fee falls in your own framework. A person sets the fee.
• Every package states what is included and what triggers extra work, so scope creep is visible later.
• Client information goes only into a business plan with training switched off, under your firm’s AI policy.

Then feed in the year’s job notes and ask where the work has outgrown the package. That is your repricing list, built from evidence. Our guide to raising prices covers the letter that follows.

Learning check

What would you do? Pick one; every answer is explained.

After a discovery call with a prospective client, an AI tool has drafted a structured scope of work and suggested where the fee falls in your firm's own pricing framework. What happens next?

1. a Someone who knows the client checks the scope and sets the fee.
2. b Discard the draft and write the quote from memory.
3. c Send the quote as drafted, since it used your framework.

A team member wants to paste a prospective client's financial statements into their personal AI account to speed up scoping. What is the best response?

1. a Fine, as they have not signed an engagement yet.
2. b Fine, as long as they delete the chat afterwards.
3. c Use the firm's business plan, with training switched off.

A client on a fixed-fee package keeps asking for extra pieces of work, and you cannot tell whether the package covers them. What should the package have stated from the start?

1. a How many hours the package is expected to take.
2. b What is included, and what triggers extra work.
3. c Nothing specific, to leave room for small requests.

Your decision for this step (saved with your progress, and in your summary)

Mark step 4 done

Some clients will ask: if AI does it, why does it cost the same? Answer before they ask.

• Say you use AIClients who find out later feel misled. Check what your engagement letter and privacy notice already say.
• Say what it changesFaster turnaround, fewer data-entry errors, more of your time on their questions.
• Say what it does not changeA qualified person reviews the work and is accountable for it. The fee pays for the outcome and that accountability, not the keystrokes.
• Point at the new workIf the package now includes something it did not, name it. That is your strongest answer.

If a client wants the cheapest mechanical service, offer one deliberately, as a narrower package. Just do not let the hourly formula drift every client there by default.

AI is the accelerator to implementing those things extremely fast.

Learning check

What would you do? Pick one; every answer is explained.

Your firm now uses AI tools to prepare parts of client work. When should clients hear about it?

1. a Before they ask: in the engagement letter and at renewal.
2. b Only if a client asks about it directly.
3. c Never, as it is just internal software like any other.

A client asks: 'If AI does the work, why does it cost the same?' Which answer is strongest?

1. a Say the AI saves very little time on their file.
2. b Offer a small discount to keep the client happy.
3. c Point to new work in their package and the qualified reviewer.

A client tells you they only want the cheapest possible basic compliance service from your firm. What is the best move?

1. a Move just this client back to hourly billing.
2. b Offer them a narrower package, chosen on purpose.
3. c Decline them, as every client needs the full package.

Your decision for this step (saved with your progress, and in your summary)

Mark step 5 done

For Australian members of CA ANZ, CPA Australia and the IPA, APES 110 already governs how you set fees, and it applies to an AI-drafted quote exactly as it applies to a handwritten one. We covered the relevant paragraphs in the ethics standard and AI pricing: the fee must support the work, and the client must understand what it covers. UK readers should look at the ICAEW or ACCA code on fees; US readers, the AICPA Code of Professional Conduct and, for tax work, Circular 230. This is general information, not legal or ethical advice.

Pricing and onboarding with AI is also one of the most-registered themes at The Firm’s AI in Practice Summit this week. “Pricing & onboarding with AI: easy for you and your clients” with Ethan Cooney (Ignition) and Ben Walker (Inspire Accountants) is on the agenda for 12 November. Notes from the summit will be at thefirm.media/summit/november-2026/notes after the event.

Learning check

What would you do? Pick one; every answer is explained.

AI has halved the time a job takes, and a partner worries it is unethical to keep charging the same fee. For Australian members of CA ANZ, CPA Australia and the IPA, fees fall under APES 110, the profession's code of ethics. What does it actually ask?

1. a Whether the fee supports the work and is clearly explained.
2. b Whether the fee matches the hours the work took.
3. c Nothing, as an AI-drafted quote falls outside the code.

A colleague who works in the UK and belongs to the ICAEW, a UK accounting body, asks which ethics code covers how they set fees. What do you tell them?

1. a Circular 230, the US tax practice rules.
2. b APES 110, the Australian code of ethics.
3. c The ICAEW Code of Ethics, which covers fees.

Your decision for this step (saved with your progress, and in your summary)

Mark step 6 done

The assumption most likely to be wrong is that clients will keep paying for accountability once AI drafts most of the work. If AI tools aimed at business owners become good enough that clients stop seeing the review as worth paying for, the mechanical end of compliance will price down whatever any firm decides. We also could not find independent data, as at November 2026, on how clients react when told a firm uses AI; the survey figures above come from a Xero-commissioned study and describe what practices intend, not what they achieved.

Should I tell clients we use AI before they ask?

Yes, in the engagement letter and in plain words at renewal. Telling them first lets you frame it around review and responsibility. Being found out frames it around the bill.

What if my fixed fees were built from hours?

Most were. Keep them while you rebuild each package from outcomes and responsibility, and stop refreshing them from new time data. Use time only to check that each package still clears your cost floor.

Can I still bill some work hourly?

Yes. Genuinely unpredictable work, such as an ATO or IRS dispute or a messy catch-up, can suit time-based billing, ideally with an estimate and a cap. The point is to choose hourly for a reason, not by habit.

Is it ethical to keep the fee when AI halves the time?

The ethics codes ask whether the fee supports the work and whether the client understands what it covers, not whether it matches the hours. Being clear about what is included, and actually delivering the review, does both.

Should an AI tool produce the quote?

It can draft one from your own framework and flag what discovery missed. A person who knows the client should confirm the scope and set the fee, because they carry the responsibility for both.