Mike Hayward, chief creative officer at Copacino Fujikado, told Digiday that clients now try tools like ChatGPT live in meetings, in front of the agency, and that it makes it harder to justify what the agency is worth. I keep coming back to that scene when I think about agency pricing models and AI, because it's the moment the old maths stops working.
Say a monthly performance report used to take an account manager six hours. With AI doing the first pull of the data and the first draft of the commentary, it now takes two. If you bill by the hour or the day, you've been handed an awkward choice: bill two hours and lose most of the fee, or keep billing six and hope nobody asks. Neither of those is a pricing strategy.
I think the question rests on a belief worth pulling apart, which is that the client was ever paying for your time in the first place.
Why hourly and day-rate billing gets squeezed
If you sell time, every hour AI saves is an hour you can't invoice. Your costs don't fall with it. Salaries and rent stay where they are, so the margin you hoped AI would give you drains out through the timesheet.
Most agencies are still priced this way. In 2025 the IPA and Creative Salon ran 63 interviews with agencies, clients and intermediaries for a report called The Price Isn't Right. They found the industry still leans heavily on time-and-resource pricing, mostly because it's easy to compare and fits how procurement teams buy. Outcome-based pricing gets talked about and rarely happens. The same release cites an IPA survey where only 27% of agencies felt they were paid a fair price.
Clients are doing the sums too. Marketing Week, reporting WFA figures, said 58% of big brands expect to pay less for their agencies where AI is used. And when Productive, which makes software for agencies, surveyed around 180 agencies for a report published in November 2025, about a third had already been asked for an AI discount and nearly half expected the question soon.
What the client is paying for
Think about what that six-hour report is for. Your client wants to walk into their Monday meeting knowing which campaigns to cut and where the money should go next. The hours were how you worked out your cost. Somewhere along the way they became the thing you sold, mostly because they were easy to count.
So when AI shrinks the hours, the value of the report stays where it was. It can even go up, because the time you got back can go into the thinking. The IPA report makes a similar point, as VideoWeek reported: a modern FTE model should account for the value AI-enabled work creates, so agencies aren't punished for getting more efficient.
Jason Cobbold, who chairs the IPA's Commercial Leadership Group, went further in Creative Salon in March. As machines take over more of the execution, he wrote, "Strategic clarity, taste, reassurance and decision-making can no longer be bundled in for free." Which sounds obvious, but it's exactly what most agencies have been giving away inside the hours.
Some of the saving probably should reach the client, mind. Copacino Fujikado still bills hourly to actuals, so its clients get the efficiency straight away, and that's a fair call if you make it on purpose. The trouble starts when the discount happens by accident, one shorter timesheet at a time.
The agency pricing models replacing the hour
The IPA's Pricing Playbook, published in March 2026, sorts the options into input-based models (retainers and mark-ups), output-based ones (deliverables and subscriptions), outcome-based ones (business performance and equity) and hybrids. For an agency of 20 to 100 people, I'd look at three models first.
Fixed-fee scopes are the easiest move. You price the deliverable, so the monthly report or the campaign has a number on it, and the efficiency is yours to keep. Munir Haddad, co-founder of Kiosk, told Digiday that clients often want a known, fixed number. A fixed scope gives them one without tying it to your timesheet.
Retainers priced on outcomes go further. Hattie Matthews, co-founder of Uncharted, described her version to Digiday: the agency is paid for its work and earns more if it helps the client hit a business goal agreed up front. Jonathan Healey of the British agency IDHL raised the catch in the same piece, which is that this kind of pricing can pull your focus towards whatever is easiest to measure.
The third is the productised service, where you sell the same defined package at the same price again and again, and improve it each time. Luquire has run a productised model since 2019, before generative AI arrived. AI makes the sums better, because every hour you take out of a fixed package goes to you.
What to tell a client who asks if you're using AI
Say yes, and say it before they ask. A WFA survey in 2024 found 80% of big brands were worried about how their agencies use generative AI, and half planned to change their agency contracts because of it. I'd expect a client who finds out on their own to assume the worst, about the price and about their data.
Then tell them what changed for them. Something like: "Yes. AI does the first pull of the data and the first draft, and someone on the team checks every number before it reaches you. The time it saves goes into the part you hire us for, which is working out what the numbers mean and what you should do next."
If they push on price, that's your cue to move the talk from hours to what they get each month and which decision it helps them make. If you're still billing hourly to actuals, the honest answer is that the bill goes down, which is why the time to change your model is before a client asks, while you still get to choose. For what it's worth, Productive's follow-up survey of 174 agencies in March 2026 found clients weren't asking for discounts any more often than six months earlier.
Where the saved hours should go
This is the bit I care about most, because saved hours don't stay saved for long. Left alone, they leak into longer meetings and extra rounds of tidying nobody asked for, so decide where they go before that happens.
I'd put strategy first. It's the work Cobbold says can't be given away any more, so make it visible. If the report now takes two hours, spend one of the four you got back on a written view of what the client should do next, and put that on the scope so it's priced. Better work comes next. AI makes it cheap to try more routes, so test three headlines where you used to test one, or check this month against last year as well as last month. The client sees that in the work, and it makes a far better case for your fee than a timesheet does.
The last place is more clients with the same team. Jake Skoric of Delta Reality told Productive that AI had made his team three to four times more efficient. It cut their utilisation, and it let them take on a lot more work without growing the team. That's where fixed fees pay off, because the same people now deliver more scopes at the same price.
Change the model before the client asks
Where I'd start is one service where AI has already cut the hours. Write down what the client gets from it, put a fixed price on that, and run it for a quarter to see what happens to your margin. You'll need to know every step to price it well, so if the process still lives in someone's head, get it written down first. None of it holds unless your team uses AI the same way every time, either, and that comes down to how much their managers back it.
Getting a whole agency team to that point, using AI well in their real client work, is what I do at Ideal and Digital, and you can see how that works here.