--- title: "Why agency prices have not moved since 2022" description: "Production got faster and agency rates held. What Forrester, the 4A’s and a UK agency survey say about who kept the saving, and what to ask a supplier." url: https://torn.studio/en/insights/why-agency-prices-have-not-moved locale: en published: 2026-10-13 --- # Why agency prices have not moved since 2022 > **Short answer:** Agency rates have held near 2022 levels because time-based billing turns every efficiency gain into supplier margin. Forrester and the 4A’s found 75% of US agencies absorbing generative AI as overhead. A survey of 73 UK agencies found 40% holding prices and delivering more work, while 11% passed the saving on. **The position:** Agency prices sit near 2022 levels because hourly billing turns every efficiency gain into supplier margin, and rebuilding that model takes an agency years. **What it means for you:** Ask a supplier to price the deliverable and the date it lands, ask what changed in their production method since 2022, and compare quotes on price per finished output. A rate card describes how a supplier is staffed; an output price describes what the work costs you. Two things are true of production work in 2026. The tools have cut the hours a deliverable takes by a large multiple, and the rate cards read much as they did in 2022. The gap between those two facts is now measured in several published surveys, and it is wide enough to be worth a buyer’s attention. ## How far did production cost actually fall? Michael Farmer has spent three decades measuring agency scopes of work, and in June 2025 he published one in full. A campaign scope of 588 deliverables was estimated at 8,644 man-hours, which at $175 an hour came to a fee of $1,512,683. Re-estimated with AI inside the production loop, the same scope came to about 1,400 hours — a saving of 84%. The fee attached to those hours falls to $241,231 for the same 588 deliverables. A month later he ran the arithmetic across a whole agency: $61.7 million of fee income from twelve clients, 84 creative staff, and AI removing 53 of those roles. Under man-hour billing that is a $15 million reduction in fee income, or 24%. Farmer’s own conclusion is that agencies should price the workload, and his warning is that procurement will do the arithmetic first if they do not. ## Why did the price stay where it was? Under time-based billing the saving has one place to go, and that place is the supplier. An agency that reports the lower hour count reports a smaller fee for identical work, so honesty about method and the revenue line pull in opposite directions. The incentive is structural, and it needs no bad actors to operate. Forrester and the 4A’s measured where the money settled. Their 2025 study of US marketing agencies found 75% absorbing generative AI as overhead, and 6% treating it as a billable line of business. The 2026 edition found 61% still classifying AI as a cost of business, 81% naming staff productivity as their main objective for it, and 31% planning to monetise agentic AI within two years. Productivity is what agencies bought; a product to sell is what most of them have yet to build. ## Who kept the saving? Mat Bennett, who advises founder-led agencies, put the question to a UK panel in September 2026. Of the 73 agencies that answered on pricing, 40% had held their prices and delivered more work for the same fee, 27% had not yet decided, 15% had raised their effective rates, and 11% had passed the saving to clients. One respondent described the mechanism in a sentence: “What we used to charge 10 days for takes 30 minutes, but we still end up doing 10 days of work for clients, just on other things.” Bennett’s reading is that the efficiency left the agency in two directions at once, towards clients as unbilled extra scope and towards model vendors as subscription and token cost, and that only 21% of his respondents had converted any of it into margin they could name. The buyer in that arrangement is getting a return. It arrives as volume, and a fee that holds is what pays for it. ## Why are buyers not asking for the difference? Pressure for a discount has stayed flat. Productive surveyed 174 agencies and consultancies between 17 and 31 March 2026 and found no significant increase in client requests for price reductions over the preceding six months. Among the agencies getting the most out of AI, 52% were holding or raising prices on improved margins. The reason shows up on the buyer’s side of the table. Agency chief executives interviewed by Digiday in July 2026 split their clients into four roughly equal groups: a quarter actively building with AI, a quarter willing but unsure how, a quarter frozen, and a quarter treating the whole thing as a fad. Three of those four groups have no working vocabulary for asking what a deliverable ought to cost in 2026. A market where the buyer cannot frame the question is a comfortable market to sell into. ## What would it cost an agency to change? Martin Sorrell has been moving S4 Capital off time and materials towards fixed annual subscriptions that bundle senior people, agentic workflows and a brand knowledge base, targeting a quarter of revenue from them by the end of 2026. The transition has been expensive. S4’s 2025 net revenue came in at £673 million, down 10.8% on the year, while the subscription line was still being built. Sorrell describes the obstacle as change management: existing contracts have to be reopened, and procurement teams are uncomfortable pricing outputs at all. One figure from the same set frames the whole question. S4 reports that spending by its technology clients on agencies has been roughly flat since 2022, while those same clients grew AI infrastructure spending by 133%. The budget moved; the agency line held. Sorrell’s own summary of where the efficiency currently lands is blunt: fifty assets a month becomes seventy, at no fee increase. An agency that wanted to price production at its 2026 cost would be rebuilding its compensation model, its utilisation targets, its client contracts and its career ladder at once, and shrinking its headcount while it did so. That is a multi-year programme with a revenue dip in the middle of it, and the firm has to keep winning work on the old terms while it runs. Agencies that start the programme late do it with less margin to absorb the dip, which is the part of the timing that compounds. ## What does a studio built after the shift do differently? A supplier that started after 2023 has no rate card to defend, no utilisation target to hit and no delivery team whose hours have to be filled, so it can quote the finished thing directly. Torn Studio sets a fixed price for each engagement before the work starts and publishes the bands, which makes the studio’s own claim checkable in the one way that matters here: the number is on the page, next to what it buys. The honest limit on that claim is capacity. One person’s calendar is the ceiling, and a bad estimate comes out of the studio’s week. What the model buys a reader is a price they can compare before a sales call, which is the thing an hourly quote is structurally unable to give them. The [published price list](https://torn.studio/en/pricing) is the place to start that comparison. **Read next** - [AI web agency or traditional agency?](https://torn.studio/en/insights/ai-website-agency-vs-traditional-agency) - [What a website actually costs in 2026](https://torn.studio/en/insights/what-a-website-actually-costs) - [When building gets cheap, deciding gets expensive](https://torn.studio/en/insights/when-building-gets-cheap) - [Fixed price or retainer?](https://torn.studio/en/insights/fixed-price-or-retainer) ## The case against The strongest argument against the position above, and what would change our mind. The hour was always a proxy for judgement, and a firm that holds its price while absorbing model costs is funding the review, the accountability and the professional risk that a low output price leaves with the buyer. Forrester and the 4A’s found 61% of agencies classifying AI as a cost of business, which is what paying for it out of margin looks like. Buyers did collect a return in the UK survey, in the form of the same fee buying more work. Evidence that incumbent price per deliverable had fallen to meet production cost, with quality held steady, would settle this the other way. ## Common questions ### Is my agency billing me for hours it did not work? The published evidence does not support that reading, and it points somewhere less dramatic. Agencies mostly estimate a scope, agree a fee and keep the efficiency as margin or as extra scope. That is a pricing choice made in the open, and the UK survey shows most agencies making it deliberately. ### How do I ask a supplier what their production method is now? Ask which parts of the deliverable are model-assisted today that were hand-made in 2022, who reviews that output, and how long the review takes. A supplier who has genuinely changed method can answer in specifics. The answer also tells you where the quality risk sits. ### Does a lower price mean lower quality? It means a different cost structure, and quality has to be checked separately. Ask for work samples produced under the new method, ask who signs off, and ask what happens when a deliverable comes back wrong. Price per output and quality are two questions, and both deserve an answer. ### Should I just demand a discount at renewal? A discount request on an hourly contract usually returns a smaller estimate for the same scope, which changes very little. Asking for the scope to be priced per deliverable with a date attached moves the risk, and it gives you something comparable across suppliers. ### Are the big holding companies actually changing? S4 Capital is moving towards fixed annual subscriptions and targeting a quarter of revenue from them by the end of 2026, against net revenue that fell 10.8% in 2025. Forrester found 31% of US agencies planning to monetise agentic AI within two years, so the intent is broad and the delivery is early. ### Why would a small studio be cheaper than a large agency? Most of the difference is coordination. A large agency prices the handovers between specialists alongside the work; a one-person studio holds the decisions in one place and prices the finished output. The trade is capacity, cover and the ability to staff several workstreams at once. ## Sources - [Nine in 10 US marketing agencies use AI to cut costs at the expense of creativity — Forrester](https://www.forrester.com/press-newsroom/forrester-nine-in-10-us-marketing-agencies-use-ai-to-cut-costs-at-the-expense-of-creativity/) — Backs the 2026 figures: 61% classifying AI as a cost of business, 81% naming staff productivity, 31% planning to monetise agentic AI within two years. - [AI is a growth engine, not a cost cutter — 4A’s](https://www.aaaa.org/blog/ai-is-a-growth-engine-not-a-cost-cutter/) — Backs the 2025 study figures: 75% of US agencies absorbing generative AI as overhead and 6% treating it as a billable line of business. - [AI made your agency faster. Who kept the saving? — Mat Bennett](https://www.matbennett.com/who-kept-the-saving/) — Backs the UK survey shares and the quoted agency response, from 73 agencies answering the pricing question in September 2026. - [The urgent case for changing creative agency remuneration before AI destroys man-hour billing — Michael Farmer](https://michaelfarmer.substack.com/p/the-urgent-case-for-changing-creative) — Backs the 588-deliverable scope: 8,644 man-hours at $175 giving $1,512,683, re-estimated at about 1,400 hours and $241,231. - [The destructive effect of AI on agency fees if man-hour billing continues — Michael Farmer](https://michaelfarmer.substack.com/p/the-destructive-effect-of-ai-on-agency) — Backs the agency-wide arithmetic: $61.7 million of fee income, 84 creative staff, 53 roles removed and a $15 million fee reduction. - [Agencies in the AI era 2.0 — Productive](https://productive.io/reports/agencies-in-the-ai-era-pulse-report/) — Backs the March 2026 survey of 174 agencies: no significant rise in discount requests, and 52% of strong AI adopters holding or raising prices. - [The AI gap between agencies and their clients is widening, not closing — Digiday](https://digiday.com/marketing/agency-bosses-say-the-ai-gap-with-clients-is-only-getting-wider/) — Backs the four equal groups of CMOs described by agency chief executives in July 2026. - [Sir Martin Sorrell says the billable hour is dying — Digiday](https://digiday.com/media-buying/its-about-change-management-sir-martin-sorrell-says-the-billable-hour-is-dying-but-getting-clients-to-move-in-is-proving-harder/) — Backs the S4 Capital subscription target, the £673 million net revenue, the flat technology-client spend since 2022 and the fifty-to-seventy assets illustration.