Everyone agrees AI saves time. Nobody agrees who keeps it.

The 2026 argument over agency fees is not about whether AI works. Both sides of the table agree it saves time. A new survey of 760 creative professionals found 81% have already shipped AI-assisted content, and the fight has moved to the invoice: who gets to keep the savings the machines created.

That question has been sitting inside every agency renewal since generative tools arrived, and two fresh data sets now put numbers on it. Luma's inaugural State of Creative Professionals and AI: 2026 report, surveyed across agencies, brands, production companies and freelancers and presented at Brandweek, asked both sides what AI should do to price. The World Federation of Advertisers ran the same question through 69 multinationals representing $147 billion in marketing spend. The answers point in opposite directions.

A balance scale with a clock on one pan and coins on the other, tilted toward the coins

Both sides want the AI savings to show up in agency fees

Luma's survey found 55% of agency creatives expect AI adoption to raise their fees, because the same team can now produce more concepts, more variations and faster turnarounds. On the brand side, 53% expect to reduce agency spending, because internal teams with their own AI tools can absorb more of the work in-house. Everybody agrees AI saves time. Agencies believe the saved time should cost clients more. Clients believe it should cost them less.

The adoption data explains part of the gap. Agencies are further along: 36% of agency creatives count themselves AI Leaders, against 17% of brand creatives. Agencies also churn through tools faster, evaluating 6.5 and dropping 3.5 in a year. The side doing more of the work believes it is selling more value; the side writing the cheques believes it is buying the same value more cheaply. Neither number is a negotiating position yet.

The labor market has already started pricing the agency case: {{link}} put a 34% hourly premium on finishing work rather than tool access, which is demand paying for judgment before any contract catches up.

The labor market has already started pricing the agency case: the AI video talent market put a 34% hourly premium on finishing work rather than tool access, which is demand paying for judgment before any contract catches up.

Video is the hardest place to prove the premium

If any category should support a fee increase, it is video, and the survey data says video is exactly where the premium is hardest to prove. No single AI task has crossed 50% adoption: 42% of creative professionals use AI for image generation and 38% for video editing assistance, the fastest-growing category as AI video tools move from novelty to production tool. Copy and stills are well past video, for the obvious reason that motion has to hold together across time.

The leaders are not winning on tool count. AI Leaders evaluate more tools, 6.5 against 2.9 for Laggards, but the real gap shows up in how much of the work AI actually touches: 78% of Leaders have released video projects where AI generated most or all of the content, against 11% of Laggards. That is a seven-fold gap in shipped work, not in subscriptions. The average team, meanwhile, evaluated 6.1 tools, adopted 4.3 and dropped 2.9 over the past year, and 52% of organizations have no structured, enforced process for evaluating or approving AI tools at all.

That churn is the attribution problem in commercial form. When a team has run six tools and kept four, nobody on either side of the table can say which tool created the value that supposedly justifies a higher fee. One in three respondents say adoption is driven by individual employees experimenting on their own. The invoice arrives before the evidence does.

The maturity gap behind the churn is well documented: {{link}} found universal AI tool access with almost no workflow maturity, and the fee fight is really an argument about who is responsible for closing it.

The maturity gap behind the churn is well documented: the AI video capability gap found universal AI tool access with almost no workflow maturity, and the fee fight is really an argument about who is responsible for closing it.

A vertical video frame with a blank price tag hanging from its corner

The contract is already moving from hours to outputs

While agencies and brands argue about percentages, the contract machinery has been moving for fifteen years. The WFA and Agency Mania Solutions study of 69 multinationals found labour-based compensation, paying for people and hours, has collapsed from 54% of arrangements in 2011 to 33% in 2022 and just 19% in 2026. Fixed-fee and output-based models climbed from 20% to 33% over the same period, and hybrid labour-plus-performance deals more than doubled from 9% to 21%.

AI is accelerating a transition that was already underway, but almost nobody has rewired the paperwork for it. Only 20% of brands have begun evolving their agency commercial models in response to AI, although 61% intend to. When AI efficiencies arrive, 33% of brands simply took reduced fees or credits, 23% accepted better quality or faster turnaround at the same cost, and only 8% shifted toward value- or output-based pricing. Meanwhile 82% agree that poor agency use of AI could make them reconsider retaining a partner: AI capability has become a roster expectation before it has become a pricing mechanism.

The transparency gap makes the fight worse. Only 48% of clients feel they have full visibility into how agency costs and profitability are calculated, and in creative work 90% say they see the value while just 41% say they see the cost. Performance pay is supposed to settle these arguments, but for 67% to 80% of respondents it still accounts for less than 20% of total compensation. Payment terms stretched too: 49% extended media agency terms over the past 18 months, and average terms now run around 75 days.

An hourglass whose falling grains become small deliverable boxes

The teams repricing well are selling proof, not tools

The report's own case studies show what a defensible AI-era fee looks like. Aperture, a growth marketing agency, tripled in size in twelve months because clients wanted more ad variants delivered faster than headcount could scale. Instead of adding another tool, the agency built AI into the workflow it already had: strategists review and approve every brief before the system generates, and humans step in whenever a shot needs a judgment call. Cost per acquisition fell more than 75% for one client, and that client raised ad spend nine-fold. The fee conversation followed the proof.

OneDay shows the other honest basis for a premium. The studio had a thirty-year-old idea on the shelf, a daredevil story with a dog riding a snowcat, long considered too expensive to make. The directors spent weeks with the original creators on characters, backstories and tone before touching a generator, then worked alongside actors, sound designers and session musicians who could direct and refine shots. The result was an 85-shot film made by a full human creative team out of an idea that had been impossible. That is expansion, not efficiency, and expansion is the only thing worth a price increase.

The failed version of the pitch is already visible in the data. Saying we use AI is becoming the creative-business equivalent of advertising that an office has Wi-Fi. With 65% of freelancers counted as AI Laggards and twice as likely to have no formal adoption process, the market is not short of access. It is short of proof that the technology improved performance, expanded what was possible, or accomplished something that was not previously practical.

Differentiation is the other half of the defense, and it starts in the briefing: {{link}} makes the case that standing out is a briefing problem, which is strategy work no subscription can do.

Differentiation is the other half of the defense, and it starts in the briefing: the sea of sameness makes the case that standing out is a briefing problem, which is strategy work no subscription can do.

What a 2027 agency contract should actually specify

The fee war ends where the contract stops counting hours. Looking ahead, 63% of brands expect to use more performance-based compensation, 46% more value-based models and 36% more fixed-fee or output-based deals. For video work, where deliverables can be defined up front, output-based pricing is the natural fit: scope the variants, the mastering spec and the approval rounds, then price the package rather than the days.

Three clauses do most of the work. First, a transparency clause that opens the cost stack, because the 48% visibility number is the real dispute and creative's 90%-value-versus-41%-cost gap is where distrust grows. Second, a sharing mechanism that states what happens when AI cuts production time: discount, more output, or reinvested strategy, decided before the savings land rather than after. Third, a proof obligation on both sides, because the teams that repriced well did it against measured outcomes, not promises.

The staffing clause matters as much as the pricing one: {{link}} puts the premium on the people who decide what ships, which is the line item an AI-era contract should protect first.

None of this resolves who was right. It relocates the argument from a percentage fight to a specification fight, which is the only version of it that produces a contract both sides can sign. Caroline Ingeborn's framing from the Brandweek presentation is the one to negotiate against: the question is no longer whether to use AI, but where AI creates value and where human judgment should stay firmly in control. The fee should follow the answer.

The staffing clause matters as much as the pricing one: rebuilding video teams around judgment puts the premium on the people who decide what ships, which is the line item an AI-era contract should protect first.

Put the framework into production

These related pages connect the article’s planning advice to a specific commercial scope.

Short-form ad productionTurn hook strategy into platform-ready creative variants.AI UGC productionBuild creator-style openings into a controlled testing system.

References

  1. How Creative Teams Are Integrating AI in 2026 (State of Creative Professionals and AI: 2026)Luma

    Luma's inaugural State of Creative Professionals and AI: 2026 report, based on a survey of 760 U.S. creative professionals across agencies, brands, production companies and freelancers and presented at Brandweek, found 55% of agency creatives expect AI adoption to raise their fees while 53% of brand creatives expect to cut agency spend; 36% of agency creatives count themselves AI Leaders versus 17% of brand creatives; teams evaluated 6.1 AI tools on average, adopted 4.3 and dropped 2.9 (agencies: 6.5 evaluated, 5.1 adopted, 3.5 dropped); 78% of AI Leaders have released AI-majority video versus 11% of Laggards; no single AI task crossed 50% adoption (42% image generation, 38% video editing assistance); 52% of organizations lack a structured process for evaluating AI tools; and case studies include Aperture (CPA down more than 75%, client ad spend up 9x) and OneDay (an 85-shot film from a 30-year-old shelved idea).

  2. Global brands are shifting to paying for outputs not time, WFA researchWorld Federation of Advertisers

    The WFA and Agency Mania Solutions report How Brands Define, Scope and Reward Agency Work, based on 69 multinational companies representing $147 billion in combined global marketing spend (fielded June 2026, benchmarked against 2018 and 2022), found labour-based agency pay fell from 54% of arrangements in 2011 to 19% in 2026 while fixed-fee/output models rose from 20% to 33% and labour-plus-performance from 9% to 21%; 63% of brands expect to increase performance-based compensation, 46% value-based and 36% fixed-fee/output; only 20% have evolved commercial models in response to AI though 61% intend to; 33% received reduced fees or credits from AI efficiencies while only 8% shifted to value- or output-based pricing; 82% say poor agency AI use could prompt them to reconsider retaining a partner; and average payment terms now run about 75 days.

  3. New Study Finds AI Has Become Mainstream in Creative WorkLos Angeles Times

    The Los Angeles Times' Brandweek 2026 coverage of Luma's State of Creative Professionals and AI: 2026 report confirms the survey of 760 U.S. creative professionals found 81% have released AI-generated or AI-assisted content and 62% are concerned about overreliance on AI at the expense of human creativity; it reports the agency economics split (55% of agency respondents expecting AI to justify higher fees against 53% of brand-side respondents expecting reduced agency spending) and quotes Luma COO Caroline Ingeborn: the question is no longer whether to use AI, but where AI creates value and where human judgment should stay firmly in control.

Related reading

AI Video Talent Market 2026: The 329% Skill Surge and the Rate Gap It CreatedThe AI Video Capability Gap: Why 2026 Teams Buy Tools and Still StallThe Sea of Sameness: Why AI Video Differentiation Is a Briefing Problem in 2026AI Video Teams in 2026: Rebuild the Org Chart Around Judgment