MAM
AI has walked into the agency, and it wants the retainer
Clients want fee cuts and the platforms want the whole job. Agencies are running out of time to work out what they are still being paid for.
MUMBAI: Bata India did not sack its agency with a press release. It simply stopped needing one.
In August the shoemaker said it had moved its entire creative production in-house, onto an AI stack built with Zocket, and was saving more than $1 million a year in agency costs. Its marketing team once briefed and chased agencies. Today it gives instructions to software, and the software does the work.
Multiply Bata by a few thousand advertisers and you have the story of the agency business in 2026. The machine is no longer a tool on the art director’s desk. It is sitting in the client’s office, doing the job the agency used to bill for.
Look at the payrolls. Employment at US advertising agencies fell 11.7 per cent between November 2022, when ChatGPT arrived, and August 2026, going by US Bureau of Labor Statistics data. In the year to August alone, American advertising, PR and related services shed 6,900 jobs.
The holding companies are leading the cull. WPP had 115,473 people at the end of 2022. By June this year it was down to 97,388, and more cuts are due before December under chief executive Cindy Rose. Omnicom, fresh from swallowing IPG, wants to get to about 105,000 staff by year-end from roughly 120,000 at the end of 2025. Its chief financial officer, Phil Angelastro, has said AI will likely push headcount lower over time.
Not every lost job is a scalp for AI. Merger overlap and a slower ad market are doing their bit too. But the stock market has delivered its verdict. Since end-November 2022 WPP’s shares have fallen 52 per cent and S4 Capital’s 73 per cent.
Publicis is the odd one out. Its shares are up 72 per cent over the same stretch and its workforce has grown to about 114,000. Chief executive Arthur Sadoun told analysts this month that AI will take out a quarter of manual tasks, not jobs. Publicis bought into data and technology years before its rivals did, with Epsilon and Sapient. It is now being paid for that call.
Job cuts are the symptom. The bigger change is in who owns the machine.
Digital now takes 68 per cent of India’s ad spend, WPP Media reckons. Google and Meta alone pocketed 64 per cent of the country’s Rs 94,700 crore digital ad market in 2025, says the FICCI-EY report. E-commerce platforms such as Amazon and Flipkart took another 23 per cent, and their share is rising fast.
Each of these platforms owns its customers and its inventory. Now it also owns the AI that writes the ad, picks the audience and sets the bid. The agency was built to stand between a brand and a scattered media market of hundreds of newspapers and dozens of channels. That market has collapsed into a handful of walled gardens, and every one of them would rather deal with the brand directly.
Indian agencies have had their pay squeezed before. For decades they lived off a 15 per cent commission on media bills. Clients broke that model, moved to fees and retainers, and media buying was split off into separate agencies in the late 1990s. Each time the business survived by finding something new to bill for. What is different this time is the speed. The last shift took a decade. This one is taking quarters.
It helps to split the agency world in two, because AI is hitting each half in a different place.
Creative agencies face a cost collapse. The work they sell is getting cheaper to make by the month. Indian brands claim media reports that AI has cut campaign turnaround by 70 per cent and production costs by 85 per cent. When the cost of making something falls that far, the price clients will pay for it follows.
Media agencies face a different squeeze. Their value lay in planning and optimising, deciding where the money goes and tweaking it every day. On Google and Meta the platforms’ own AI now does most of that tweaking inside a black box. A media planner cannot beat Advantage+ at its own game when she cannot see inside it.
So the big media groups are changing what they sell. They increasingly buy inventory in bulk as principal and resell it to clients at a markup they do not always disclose. Omnicom’s third-party service costs, which include media it buys as principal, jumped to almost $2.9 billion in the first half of 2026 from $1.7 billion a year earlier, Digiday reported. Some holding companies are even throwing in free AI tools for clients who sign up for principal buying.
It is good business, but it changes the relationship. An agent works for the client. A trader works for its own margin. Independent agencies argue that a firm buying and selling the same inventory is no longer an agency at all. Indian advertisers, who squeeze every paisa out of a media plan, will ask the same question sooner or later.
There is an Indian twist too. The muscle of Indian media agencies was built on buying television, where scale and hard bargaining still counted. WPP Media’s midyear forecast has TV ad revenue falling 6.8 per cent this year. The one medium where humans could still haggle is the one that is shrinking.
In June 2025 The Wall Street Journal reported that Meta wanted brands to be able to create and target ads entirely with AI by the end of 2026. Agency stocks fell between 2.9 per cent and 4 per cent that day.
The deadline is two months away and Meta is close. Its Advantage+ tools already turn still product shots into video ads. At Advertising Week this month it said it was testing an assistant that creates campaigns and shifts budgets on a chat command.
Now for the sting. Meta’s revised commercial terms, which kick in within weeks, make advertisers solely responsible for whatever its AI creates or changes in their ads, including features Meta switches on by itself. US agencies such as Exverus have already pulled live ads after Meta’s AI added links, translations and images nobody had asked for. The platform builds the ad and the brand carries the can.
Then there is the new gatekeeper. OpenAI switched on ChatGPT ads in India on 27 August, with WPP and Omnicom as launch partners. A week later it opened a self-serve platform with daily budgets starting at Rs 725.
Agencies should recognise the pattern. Google courted them too when search advertising was young, then built tools good enough for millions of small advertisers to skip them altogether. A new platform needs agencies at launch to bring in the big spenders. It needs them a lot less once the self-serve tools work.
Agencies have always sold time. A retainer is a headcount in a suit. Juniors resize banners, cut versions, write fifty social posts and build media plans line by line. Clients pay for those hours, and the hours fund the senior people and the big idea.
AI goes straight for that base of the pyramid. When 200 versions of a campaign can be churned out over a lunch break, nobody will pay for a fortnight of studio time. Pull out the production hours and the whole business model starts to wobble.
There is a hidden bill here that few agency bosses are adding up. Junior roles are where the business trains its future leaders. The account executive resizing banners today is the planning head of 2035. eMarketer notes that Americans aged 20 to 24 have largely vanished from the industry and that internships keep falling. Cut the bottom rung now and there will be nobody worth promoting in ten years. Agencies that teach their juniors to direct the machines, rather than replacing them with machines, will own the next generation of talent.
Indian agency bosses can take some comfort from the topline. WPP Media expects India’s ad market to grow 8.8 per cent this year to around Rs 2 lakh crore. That is far healthier than anything in the West.
A bigger pie does not mean a bigger slice for agencies, though. Indian marketers have been demanding fee cuts of as much as 25 per cent. Media reports argue that if AI cuts design or media management time by a quarter, fees should fall by a quarter too. Adani Group general manager, digital media, Chandan Sharma admitted that agencies have been pushed into signing for lower fees.
The sharper threat is coming from India’s own talent pool. Global companies such as Kimberly-Clark, Target and Catalyst Brands are using AI at their Indian capability centres to pull advertising and creative work in-house, Reuters reported. The Bengaluru engineers who might once have serviced an agency account are now building the machine that replaces it.
Most Indian agencies still bill on manpower-based retainers. That is precisely the model under fire.
The rulebook is changing as well. On 29 September the Advertising Standards Council of India (ASCI) issued guidelines on labelling synthetically generated content in ads, which take effect three months after publication. A label will be needed wherever AI content could materially sway a buying decision. Someone has to police that, ad by ad.
There is an upside if Indian agencies are quick enough to grab it. The country already runs the back offices of global brands and networks. AI-run production in English and a dozen Indian languages, at Indian costs, is an export business waiting to be built. The capability centres have spotted it. The question is whether the agencies get there before the clients do.
Groups with their own data and technology, Publicis being the obvious case, are already being rewarded by clients and investors. Media groups with big trading desks will keep making money, at least until clients start asking hard questions about markups. Small, AI-native independents can undercut on price and move faster than any network.
The squeeze falls hardest on the middle. The mid-sized creative shop that sells hours, owns no data and runs no trading desk has nowhere to hide. In India, that describes a large chunk of the business.
Sell outcomes, not hours. Omnicom has already spoken of a shift from paying for people and hours to paying for outputs and results. Indian agencies will have to follow, and quickly. A client who can see AI’s savings on a spreadsheet will not keep paying a manpower bill.
Run the brand’s AI control room. Brands will increasingly operate the platforms’ machines themselves. Somebody still has to feed those machines strong brand assets and check what comes out before ASCI or the lawyers do. Somebody has to stop Meta’s AI slapping a clumsy translation on a luxury brand. That is skilled, billable work, and agencies are well placed to do it.
Get found by the bots. Consumers are asking ChatGPT and Gemini what to buy instead of scrolling past ads. Getting a brand recommended by an AI assistant is a new trade, and Dentsu India has already expanded its generative engine optimisation practice. Most large agencies will need a similar unit within a year.
Come clean on trading. If media agencies are going to make their money as principal buyers, they should say so and price it openly. The ones that hide the markup will win a few quarters and lose the client’s trust for good.
Shrink the pyramid, fatten the top. The model of hundreds of juniors feeding a handful of creative stars is on its way out. Agencies will be smaller and more senior, with more revenue per head. US consultant Michael Farmer has argued that AI savings should be ploughed into senior thinking rather than deeper fee cuts. Agencies need to make that case to clients, with hard numbers.
Guard the idea. AI can knock out a thousand ads. It still struggles to spot the cultural moment that makes one of them stick in the memory. That is where an agency can still charge a premium.
Two years ago agencies were asking whether AI would take their jobs. Bata and a growing list of clients have answered that for them. The question now is what agencies have left to sell once the hours are gone.
Those that find a good answer fast will live to pitch another day. The rest will end up as a single line on a client’s cost-savings slide.




