Every major agency network has launched an AI practice in the last eighteen months. Most of them are selling capability decks. Almost none of them are selling capability. The brands buying these services are about to find out.
A CMO I cannot name received the AI transformation deck from her holding-company agency last quarter. Sixty-four slides. Maturity model. Phased rollout. Centers of excellence. Workshop deliverables. The deck cost €120,000. The deck was beautiful. The deck was a deck.
She asked one question. Has anyone here actually shipped AI work? The room went quiet.
This is happening at every level of the industry. The agency model is responding to AI the way it responded to digital in 2002 and social in 2008. By selling the thing before understanding the thing. The transformation deck has replaced the campaign deck. Same shape. Same logic. Same gap between what is promised and what gets shipped.
Agency AI services are mostly an organisational problem in a creative costume.
The structure that built broadcast advertising, account management, creative review, production, planning, is precisely the structure that cannot ship AI work. AI work is iterative. It is technical. It is collaborative across disciplines that do not share vocabulary. It requires people who understand both the creative idea and the system that produces it. Most agencies have separated those people for thirty years.
The credentials problem is real. The agency that won three Lions in 2023 for traditional brand work is not, by virtue of those Lions, qualified to ship a campaign generated and produced through AI tools. Different muscle. Different infrastructure. Different talent. Different process.
What is being sold instead is reassurance. Strategy decks. Frameworks. Roadmaps. Pilot programs that never become production. Workshops that produce decks that produce more workshops. The agency takes the client's anxiety about being left behind and converts it into invoices. This is good for revenue. This is bad for outcomes.
The thing AI work actually requires is rare and not easily fakeable. It requires creative direction that has actually directed AI work. Not theoretical. Not "I oversaw a pilot." Actually directed. Made the calls. Shipped the asset. Knows what fails and why.
This is a small population. Most of the people who can do this work are not at the major agencies, because the major agencies were not where AI work was happening for the last four years. The work was happening in small studios, individual practitioners, a handful of brands brave enough to try things, and the artists and directors who made AI tools their actual medium long before any agency had a Head of AI.
The agencies are now trying to hire that talent at scale. They are losing because their structure makes the work harder, not easier. The talent that can do this work knows it. The talent that cannot do this work but can pass an interview is happily taking the job.
There are three groups selling AI services right now, and brands need to learn the difference quickly.
The first group is consultants who have rebranded as AI experts. Their value is in the deck. They produce roadmaps, maturity assessments, governance frameworks, change management plans. None of this produces work. All of it sounds productive in a board meeting.
The second group is agencies that have hired a few practitioners and built capability in pockets. This is real but uneven. The same agency might ship genuinely interesting work with one team and pure deckware with another. Brands can get burned assuming the brand name guarantees the practice.
The third group is creative-led shops, small studios, and individual creative directors who have actually been making AI work for years. This is the smallest group and the most fragmented. Most brands cannot find them through normal procurement processes. The procurement process itself selects for the first group.
The brands that figure this out early will be the brands that ship interesting AI work. The brands that go the procurement route will get the deck and the invoice.
Most agency AI services are selling you the workshop, not the work.
The transformation deck has replaced the campaign deck. Same shape. Same gap.
The talent that can do this work is not at the major agencies, because the major agencies were not where this work was happening.
Ask for shipped work.
Not a case study about strategy. An asset. A campaign. A product. Look at it. Decide if it is good.
Ask who specifically directed it.
Get a name. Look up that person. Check if they have a body of work. Check if they were doing this work in 2022.
Ask what failed.
Anyone who has actually made AI work has stories about what didn't work. The story is more useful than the success.
Ask what they would do differently this time.
The answer reveals whether the practitioner has actually evolved their thinking or whether they are repeating talking points.
Ask to talk to that practitioner directly, not the account director.
If the answer is "let me set up a workshop," you have your answer.
The agency-AI gap is going to become visible in the next eighteen months. Some brands are going to spend a lot of money on capability decks and end up with nothing to show. Other brands are going to figure out the alternative path. Working with smaller, faster, more practitioner-led teams that can actually ship.
The category that is going to win is the one that does not yet have a name. Not agency. Not consultancy. Not AI shop. Something that combines creative direction, technical fluency, and the speed of a studio with the strategic depth of a partner.
This category is being built right now, mostly by people who left the agencies because they could see this coming. The brands that find them early will get the work. The brands that wait for the agencies to figure it out will pay for the wait.
The honest version of the AI transformation conversation is shorter than the sixty-four-slide deck.
If your agency cannot show you AI work they have actually shipped, by people they can name, with stories about what failed, they are not your AI partner. They are selling you reassurance.
The work is real. The market is real. The rare thing is the people who can actually do it.
Find the people. Skip the deck.