Skip to main content
Navigation

Billing for work AI sped up: let waiting work pick a rate cut or fixed fee

Bill for work AI has sped up by whether waiting work can fill the freed time, not by the size of the speed-up.

Key points

  • Is work waiting? That decides, not speed
  • Rate cut or fixed fee: cut if work waits
  • Handoff and checks stay. Price them in
  • Write the client note with the decision
Bill for work AI has sped up by whether waiting work can fill the freed time, not by the size of the speed-up.
A diagram of this article’s key points.
Contents

The deciding factor is waiting work, not the speed-up

A&A perspective

When AI shortens the work on a deliverable you have always quoted by the hour, how to bill for it is settled not by how many hours turned into how many minutes but by whether you already have work in hand to fill the freed time. In this article, waiting work means jobs you could start as soon as your hands are free.

A&A perspective

Quotes you turned down, inquiries you have left unanswered, and clients you have kept waiting all count.

A&A perspective

Automate & Augment (A&A), which rebuilds operations around AI and takes the work through to implementation, took its cue for this way of deciding from two overseas cases.

From the sources

One of the customer stories Anthropic has published describes Layer, a 25-person advertising agency in Norway.

Anthropic ↗

From the sources

Layer produced a report that pulled numbers from five different sources. It quoted its client 10 hours of billable work for one report, billed by the hour. Once it used MCP, a connector that lets Claude read data from ad platforms directly, the case study says the report came to take 20 minutes.

Anthropic ↗

From the sources

The other case is EvenUp, which builds document-drafting AI for US law firms that handle personal injury cases. Before EvenUp, those firms drafted the documents by hand, at 8 to 15 hours per case, and the case study says they were left with two options: take the case and absorb the hours, or refer it out to another firm.

Anthropic ↗

From the sources

With EvenUp, the case study says, this drafting shrank from hours to minutes. One firm that adopted it had a 45-day backlog of cases. What the case study reports as the result is not a lower price but that the number of demand letters the firms sent out tripled, and that the backlog was cleared.

Anthropic ↗

A&A perspective

A&A reads the two cases this way. On Layer's side, the very hours it billed against disappeared. Layer's case study does not say that work was waiting to fill the freed time.

A&A perspective

On EvenUp's side, cases were already queued, so the freed time went straight into the next job. What separated the two was not the size of the speed-up but whether the work was there first.

A&A perspective

Neither case study says anything about how to set prices, though. The figures are self-reported by the AI provider and the companies that adopted it, with no case count and no time period. Tripled output and a 45-day backlog are not numbers you can use for your own forecast. All this article borrows from the cases is one distinction: after a speed-up, you are in one of two situations.

Cut the rate if work is waiting; if not, move to a fixed fee per deliverable

A&A perspective

If you have waiting work, you can lower your rate and make it back on volume. If you keep billing by the hour without it, every gain in efficiency shrinks your revenue. When there is no waiting work, what you change is not the price but the unit you bill in.

A&A perspective

A fixed fee per deliverable sets a price and an acceptance standard for each deliverable rather than for the hours worked. Once the unit is the deliverable, the bill does not move when the work gets faster.

A&A perspective

There are four billing options. Waiting work decides between two of them: a rate cut and a fixed fee. The other two depend on different conditions, but they sit in the same table. All four are A&A's own design, not something drawn from the sources.

From the sources

Widening the deliverable has a close real example. At Layer, the case study says, annual media plans, once too time-consuming to build thoroughly, are now standard practice. The case study does not say whether those plans were billed separately.

Anthropic ↗

A&A perspective

A&A reads this as freed time flowing into a deliverable that could not be made before.

A&A perspective

A&A sees keeping quiet as usable only as a stopgap. Once you have quoted on the basis of hours, the client will eventually see that the basis is gone. If you have no explanation when they do, every bill you sent before may come into question.

A&A perspective

If you choose the rate cut, how to read the economics after the cut is covered in a separate article, “Low-ticket service profit rests on human touches, not AI speed.”

Four ways to bill after a speed-up (A&A's design, not drawn from the sources)
Billing optionChoose it whenDo not choose it when
Lower the hourly rateTurned-down work has piled up, and a lower rate would let you take on more jobsInquiries are not yet steady
Keep the price, widen the deliverableThe client clearly wants something moreYou have not checked with the client what to add
Switch to a fixed fee per deliverableThe deliverable has a set form and its acceptance standard can be written downSpecs change every time and the workload cannot be predicted
Change nothing and keep quietOnly as a stopgap until the next quoteThe client could spot the speed-up first, and you have no explanation ready

Handoff and checks remain after the speed-up, so keep them in the price

A&A perspective

Whichever option you choose, include the time for handoff and checking in the price. What AI mainly sped up is the making. Handing the work over and having a person check it are still there.

From the sources

In Layer's case, 17 of the 20 minutes go into moving the finished data into the client's spreadsheet. The case study also says that, before rolling it out to clients, a Layer strategist spent weeks validating the AI's output.

Anthropic ↗

From the sources

A wrong number in a client report is expensive, Layer's case study says, because someone spends real budget on it. EvenUp's case study likewise says every draft still ends with a person: an attorney reads, checks and signs it.

Anthropic ↗Anthropic ↗

A&A perspective

A&A reads this as where the split between people and AI falls. The making can go to AI, but fitting the work to the client's format and checking it while carrying responsibility for errors still sit with people. Price only the making, and those two steps become unpaid work.

Hypothetical example

Take a hypothetical. A four-person research firm quotes a client's monthly report as half a day of work and bills it at an hourly rate. With AI, the part that produces the report now takes 20 minutes. This 20 minutes is a made-up figure, separate from Layer's case.

Hypothetical example

But when the firm times checking the numbers, pouring them into the client's format and rereading before sending, those come to 90 minutes in total. Half a day did not become 20 minutes; it actually became just under two hours.

Hypothetical example

Next, the firm counts the quotes it turned down in the past three months because it could not meet the deadline, and the inquiries it has left unanswered. Both come to zero. With no waiting work, a rate cut is off the table. The firm switches to a fixed fee per report, based on the just-under-two hours that include checking and handoff.

Hypothetical example

Had it turned down three jobs, lowering the rate to win more of them would also have been an option.

Write the client note at the same time you decide how to bill

A&A perspective

Whichever of the four you choose, write the note to the client at the same time as you decide the billing form. If you leave it for later and the client notices the speed-up first, you end up explaining on the back foot. In the note, say what changes and what stays the same in the same sentence.

Hypothetical example

An example for a rate cut: “Our working time has come down, so we are lowering the rate. We are also shortening the turnaround you have been waiting on.” Give the reason for the cut together with what the client receives.

Hypothetical example

An example for widening the scope: “The price stays the same, and the summary table we used to quote separately is now included.” Put both what is included and what is not in writing. If you do not, the widened scope will be read as having no limit.

Hypothetical example

An example for moving to a fixed fee per deliverable: “From the next one, we will price each report rather than the hours worked. However the work is done inside, what you receive and the standard it has to meet stay the same.”

A&A perspective

A&A thinks that when you set the change of unit next to the unchanged standard, the client can see that what they receive has not changed.

A&A perspective

Even if you choose to keep quiet, you need a note. It is the sentence you will use at the next quote, when you move the basis of the bill from hours to something else. Keep the price without one ready, and the first explanation you give will be when the client asks.

A&A perspective

The steps for telling existing clients about a price change are covered in a separate article, “Existing customer price rise: start date, proration, failed payments.” This article covers the step before that: deciding what form the bill takes.

Summary

A&A perspective

The deciding factor is not the speed-up but whether waiting work can fill the freed time.

A&A perspective

With waiting work, cut the rate; without it, move to a fixed fee per deliverable.

A&A perspective

The time for handoff and checks remains after the speed-up, so include it in the price.

A&A perspective

Prepare the client note at the same time as you decide the billing form.

Bill for work AI has sped up by whether waiting work can fill the freed time, not by the size of the speed-up. All this article borrows from the two cases is the distinction between the two situations after a speed-up. The way of choosing a billing option is A&A's own design, not something drawn from the sources.

FAQ

Isn't it more honest to pass the AI savings on to the client as a price cut?

If you have waiting work, a price cut is a sound choice. Cut without it, and every gain in efficiency shrinks your revenue. A&A thinks honesty shows less in the amount than in updating the basis of the bill and explaining it.

Can I move to a per-deliverable fee under a contract that settles by itemized hours?

Sometimes you cannot. With clients or contracts that require an hourly breakdown, the options left are a rate cut or widening the scope. Even then, put the time for checking and handoff in the breakdown and bill for it.

What exactly should I count as waiting work?

Count the quotes you turned down in the past three months for lack of time or capacity, the inquiries you have left unanswered, and the clients you are keeping waiting now. If the count is zero, the freed time will not be filled by the next job; it will simply sit empty. This way of counting is A&A's own design, not a standard from the sources.

Sources & editorial note

Primary pages read for this article. Publication dates below belong to the sources; access dates record our research.

  1. Supermetrics lets marketers manage ad campaigns from a conversation with Claude

    Anthropic · n.d.

    Accessed 2026-10-21
  2. EvenUp cuts document drafting from 15 hours to 15 minutes with Claude

    Anthropic · n.d.

    Accessed 2026-10-21

Editorial policy

← All articles