The Legacy Ad Machine Was Built for a World That Is Ending
For thirty years, two kinds of company sat between advertisers and audiences. Agency holding companies (WPP, Dentsu, Omnicom, Publicis, IPG) bundled strategy, creative, and media buying, and monetized human hours and media-buying scale. Ad-tech intermediaries (Criteo, The Trade Desk, and the programmatic stack) monetized data and the cookie, following users around the web to retarget them. Both models made sense in a world where reaching people required either privileged access to media or a stored profile of who they were.
AI-era advertising breaks both premises at once. Ads inside ChatGPT, Gemini, Copilot, and Perplexity are bought self-serve, so you do not need an agency’s media desk to place them. They are matched by conversational context, not by a cookie, so the data-broker layer loses its reason to exist. And they demand on-brand creative at machine volume for every surface, which billable human hours cannot supply affordably. The legacy machine is not merely too expensive for the new medium, it is structurally the wrong shape for it. You cannot fix a shape problem with a discount.
The tell is in how the incumbents describe their own futures. When the growth plan of the world’s largest agency is to become smaller, merge divisions, and relabel itself “AI-enabled,” that is not a company leading a transition. It is a company managing a decline while trying to catch up to a model that challengers were built around from day one.
36% by 2028
share of all marketing work projected to be automated by AI by 2028, up from about 16% in 2026, the direct pressure on a headcount-billed model
The Numbers: What Is Happening to the Holding Companies
This is not a forecast, it is already in the financials. In its 2025 results WPP, once the world’s largest ad group, reported revenue down roughly 8% on a reported basis to about £13.55 billion and launched “Elevate28,” a multi-year plan to collapse its holding-company structure into a single “AI-enabled” business, targeting £500 million in annual cost savings and a return to growth no earlier than 2027. Dentsu booked a ¥327.6 billion statutory net loss driven by a ¥310.1 billion impairment on its international assets, grew organically just 0.5%, suspended its dividend, pulled back from M&A, and is cutting 3,400 international roles, about 8% of that workforce. Its 2026 guidance is 0 to 1% organic growth.
Read together, these are the symptoms of a model under structural strain. Impairments admit that assets bought in the old world are worth less in the new one. Restructurings admit the cost base no longer fits the revenue. And “AI-enabled” turnaround plans admit that automation is now a survival requirement rather than an efficiency bonus. When your headline strategy is to shrink your way back to health, you are not competing for the future, you are defending the past.
| Legacy player | What its own recent results show |
|---|---|
| WPP (agency holding co.) | Revenue down ~8% to ~£13.55B; “Elevate28” restructuring; £500M cost cuts; growth not targeted before 2027 |
| Dentsu (agency holding co.) | ¥327.6B net loss; ¥310.1B impairment; 0.5% organic growth; dividend suspended; 3,400 roles cut |
| Omnicom + IPG (merged holding co.) | $13.5B all-stock merger closed Nov 2025; ~$25B combined revenue; legacy brands retired; thousands of roles targeted; ~$750M cost savings goal |
| Criteo (retargeting ad-tech) | FY 2025 revenue ~$1.9B, down ~1%; retargeting fell to ~40% of the business; lost its largest retail-media client’s managed services |
The Omnicom-IPG Merger: Consolidation Is a Symptom, Not a Cure
In December 2024, Omnicom announced it would acquire Interpublic Group. The deal closed on November 26, 2025, after regulators in the US, UK, and EU signed off, creating the world’s largest advertising and marketing holding company with pro forma revenue above $25 billion. The famous “big six” holding companies (Omnicom, IPG, WPP, Publicis, Dentsu, Havas) became five overnight. On paper it looks like strength. Read closely, it is a defensive move.
Consider the mechanics. IPG had already cut roughly 3,200 employees and shed 730,000 square feet of office space before the merger even closed. Omnicom set a target of reducing staff costs by about 10% and roughly $750 million in annual savings, and one analyst estimated the combined company could shed as many as 20,000 roles as back-office functions, data platforms, and duplicate agency brands are consolidated. The stated logic is that being bigger lets the combined group cut costs and use its collective client spend to negotiate better rates from media owners and platforms. That is precisely the point: the merger doubles down on the two things AI is dismantling, media-buying leverage and human scale. As one former CMO put it, Omnicom is “betting on the future of advertising not actually being about advertising” but about data, commerce, and retail media, an admission that the traditional business is not where the growth is.
Mega-mergers happen when an industry runs out of organic growth. You combine two shrinking businesses to buy time and cut cost, not to invent a new category. The AI-native challengers do not need to merge to find scale, because software scales without headcount. That is the difference between managing decline and building the next thing.
Big six to big five
the Omnicom-IPG merger removed an entire holding company from the top tier, with up to 20,000 combined roles estimated to be cut, a consolidation driven by cost, not growth
Why the Holding-Company Model Does Not Fit AI-Era Advertising
The holding-company model is a labor-arbitrage business: sell senior expertise, staff the work with cheaper juniors, bill the hours, and use scale to negotiate media discounts. Three forces make that model obsolete in the AI era. First, creative is being automated. The junior hours that staffed campaigns are exactly the work generative AI now does in minutes, which is why Gartner projects AI will handle 36% of marketing work by 2028. Second, buying is self-serve. OpenAI’s Ads Manager lets a founder place a ChatGPT campaign without a media desk, dissolving the distribution leverage agencies sold. Third, the medium rewards speed and volume, meaning dozens of conversation-matched creatives per campaign refreshed weekly, which a retainer-and-timesheet model cannot deliver at a competitive cost.
The uncomfortable truth for incumbents is that their scale, once a moat, is now a liability. Tens of thousands of employees and a portfolio of acquired agencies are a fixed cost base optimized for a world where advertising was slow, human, and relationship-gated. AI-era advertising is fast, generated, and self-serve. You cannot restructure your way from one to the other; you have to be built for it. Every layoff and merger is an attempt to shrink the fixed cost base fast enough to survive a revenue base that is shrinking faster.
Criteo and the Ad-Tech Middlemen: A Structural Problem
Criteo is the clearest case of an intermediary whose foundation is eroding. It pioneered cookie-based retargeting, chasing a user with ads for the shoes they browsed, and for years that was the definition of “performance media.” By its own account, retargeting fell from the majority of its business to about 40% as privacy changes from Apple and Google chipped away at the cookie. Full-year 2025 revenue was roughly $1.9 billion, down about 1%, and it lost the managed-services relationship with its largest retail-media client. Even the reprieve of Google keeping third-party cookies in Chrome only slows the decline; it does not change the direction of travel, because AI assistants are creating a new, cookieless, context-matched surface where the retargeting middleman has no role to play.
The deeper issue is that ad-tech intermediaries add a layer of fees and tracking between advertiser and audience, a layer AI-native advertising simply removes. Inside ChatGPT, the platform matches your ad to the conversation directly; there is no cookie to buy, no user to follow across the web, no demand-side platform to arbitrage. The middle disappears, and the value moves to the two ends: the platform that owns the conversation, and the layer that produces the creative. For the full picture of this new stack, see LLM ad infrastructure explained and whether retargeting is dead.
The In-Housing Wave: Clients Are Already Voting With Their Budgets
Even before AI, brands were quietly walking away from the agency model. According to the Association of National Advertisers, the share of large marketers running an in-house agency climbed from 42% in 2008 to 58% in 2013, 78% in 2018, and 82% by 2023. Among those companies, an average of 61% of all creative and media work is now done internally, and 65% had moved established business away from an external agency to their in-house team in the prior three years. Media buying, long considered too complex to bring in-house, is falling too: 54% of in-house agencies now handle some media planning or buying.
The motivation has shifted in a way that should alarm holding companies. In the ANA’s 2026 research, only 9% of in-house teams cited cost savings as their main reason for existing, down from 30% in 2023, while 53% now describe themselves as strategic partners and 58% sit in upstream strategy. Translation: clients are not just insourcing to save money, they are insourcing because they believe they can do the work better and faster than the agency. Now add AI tools that let a three-person in-house team generate a full multi-platform campaign in minutes, and the last argument for the agency (that it has capacity and craft the client lacks) weakens further. The agency’s customers are becoming its competitors.
82% in-house
of large marketers now run an in-house agency, up from 42% in 2008, with 61% of work done internally, and only 9% citing cost savings as the main reason
The Publicis Exception (and What It Proves)
Not every incumbent is declining equally, and the exception proves the rule. Publicis grew organically about 5.6% in 2025 and guided to 4 to 5% for 2026, outperforming peers, because it treated AI as a growth engine rather than a cost-cutting tool, rebuilding its offering around data and automation years earlier through its Sapient acquisition and CoreAI platform. The lesson is not “big agencies are fine.” It is the opposite: the only incumbents holding up are the ones aggressively becoming AI-native, and even they are competing on the same terrain as AI-first challengers. If the defense against disruption is “become the disruptor,” the disruptors already have the structural advantage of having started there, with no legacy cost base to dismantle first.
The Fee Stack: Where the Money Actually Went
To see why the AI-native model wins on economics, look at where a legacy advertising dollar went. A meaningful slice never reached the audience at all; it paid for the layers in between.
| Legacy layer | What it charged for | AI-native replacement |
|---|---|---|
| Creative agency | Designer and copywriter hours, retainers, revision cycles | Generative creation from a prompt in minutes |
| Media agency | Media planning, buying, and negotiated discounts | Self-serve ad managers with contextual matching |
| Ad-tech intermediary | Cookie data, retargeting, DSP arbitrage fees | Platform-native contextual matching, no cookie |
| Holding-company overhead | Corporate layers, real estate, M&A integration | Software that scales without headcount |
What Replaces Them: The AI-Native Advertising Stack
The advertiser of 2026 does not assemble a holding company and a programmatic stack. They assemble a lean, AI-native stack: a platform that owns the ad inventory (OpenAI, Google, Meta), a creative-and-campaign layer that generates on-brand, context-matched ads at volume for every surface, and lightweight measurement to close the loop. The expensive, slow, fee-laden middle is gone. This is the same pattern that has played out in every industry AI has touched, where value migrates from intermediaries to the ends of the chain, and it is why the practical question for most businesses is no longer “which agency” but “which AI-native platform.” We cover the buyer’s side of that decision in do you still need an ad agency in the AI era.
How Lapis Fits the New Model
Lapis is built as the creative-and-campaign layer of that AI-native stack. From a single prompt, it generates production-ready ads for ChatGPT plus Meta, Google, Reddit, and LinkedIn in under three minutes, on-brand automatically, because Brand Intelligence learns your logo, colors, typography, and voice from your website. It replaces the exact functions the legacy machine billed for: the designer and copywriter (creative generation), the media planner (multi-platform, self-serve output), the analyst (Performance Forecasting and Web Analytics), and the competitive-intelligence team (Competitor Tracking). What took an agency a retainer and two weeks takes a founder a prompt and three minutes, at any budget.
This is not a cheaper agency; it is a different shape entirely, meaning self-serve, generated, and neutral across every channel and model provider. That neutrality matters: unlike a holding company tied to media-buying relationships or an ad-tech firm tied to the cookie, Lapis produces creative for whatever surface wins attention next, so it keeps working as Gemini, Copilot, and Perplexity open their inventory. It is the layer that lets any business participate in AI-era advertising without the legacy overhead.
3 min vs. weeks
time to produce a full multi-platform, on-brand campaign with Lapis versus an agency retainer and briefing cycle
Getting Started
You do not need to wait for a holding company to finish its multi-year restructuring to advertise in the AI era. Paste your website URL into Lapis, describe an offer in one sentence, and watch it produce a full set of on-brand, context-matched ads for ChatGPT and every other channel, with forecasts attached, in the time it used to take to brief an agency.
Start with Lapis free (5 credits, no credit card). Lapis is one of the fastest-growing Y Combinator startups (F25), rated 5.0 on G2, with more than 10,000 campaigns generated across 30-plus industries, and it is building the AdSense for the AI era: the creative and campaign layer that replaces the legacy ad machine with something any business can run.
Related guides
- Is Retargeting Dead?: the cookieless, context-first shift
- Do You Still Need an Ad Agency in the AI Era?: the practical decision
- AI Agents That Run Marketing End-to-End: what replaces the agency workflow
- LLM Ad Infrastructure Explained: the new advertising stack
- The AdSense for the AI Era: the layer that democratizes AI advertising
