OpenAI Told Every Advertiser to Fire Their Agency. Then Their Own Ads Failed. What This Means for the Future of Agency Work.

OpenAI just told every advertiser to fire their agency.

Not in those exact words. But close enough.

Asad Awan, OpenAI's head of monetisation, publicly described a future where advertisers skip agencies entirely and prompt ChatGPT to run their campaigns. His exact words: "You could say 'my goal is sell these shoes more in Midwest and go.'"

No strategy. No creative brief. No media plan. No agency. Just a prompt and a budget.

OpenAI is not positioning ChatGPT as a tool for agencies. They are positioning it as a replacement for agencies. They are approaching brands directly, offering AI-generated campaigns from concept to delivery, all inside ChatGPT.

Then the whole thing fell apart.

What Actually Happened With ChatGPT Ads

In March 2026, the first reports from ChatGPT's advertising partners started leaking. And the picture was not pretty.

ChatGPT's first advertisers cannot prove the ads work. Attribution, the ability to track whether an ad actually led to a sale, is fundamentally broken inside the ChatGPT interface. Advertisers are spending money with no reliable way to measure return.

The minimum commitment is $200,000. For context, a mid-size agency client might spend $50,000-$100,000 per quarter across all channels. OpenAI is asking for double that just to test a product with no proven attribution.

The CPM, cost per thousand impressions, is $60. Meta charges roughly $20. Google charges between $10-$30 depending on the format. OpenAI is charging three times the industry rate for an unproven ad product.

And then the user revolt happened. OpenAI rolled out promotional placements inside ChatGPT and paying subscribers pushed back hard. People who pay $20/month for a productivity tool did not sign up to see ads. OpenAI had to pull back.

CNBC reported that the ad pilot was "frustrating" partners. WinBuzzer reported that advertisers could not prove the ads work. The product that was supposed to make agencies obsolete could not clear the most basic bar: show me the numbers.

Why the Failure Does Not Matter

Here is the part most agency owners will get wrong.

They will look at OpenAI's failed ad product and feel relieved. "See? AI cannot replace us." They will use this as evidence that agencies are safe.

They are wrong.

OpenAI failed on execution. Not on intent. The vision, advertisers prompting AI directly, no agency in the loop, is exactly what every major platform is building toward. OpenAI just got there first and stumbled.

The next attempt will be better. And the one after that. And eventually, one of them will work.

The Platform Convergence

OpenAI is not the only company building this future. The convergence is happening across every major platform simultaneously.

Google launched Gemini Advantage on March 23, 2026. Their AI is now embedded across the entire Google Marketing Platform: targeting, creative, measurement, optimisation. All powered by Gemini. The pitch to brands: consolidate your tech stack inside Google. Reduce third-party layers. Activate your data in real time. "Third-party layers" is Google's polite way of saying "agencies."

Google is also auto-applying AI voiceovers to Performance Max video ads as of March 20. Opt-out, not opt-in. The platform is now creating creative assets: the one thing agencies said AI could not replace.

Meta is building toward fully AI-generated ads by end of 2026. Brands upload a product image. Set a budget. Meta's AI handles everything: creative, copy, targeting, optimisation. Everything a media buying agency currently charges for.

Salesforce renamed Marketing Cloud to "Agentforce Marketing." The word "marketing" was literally replaced with "agent." Autonomous AI agents now handle audience segmentation, campaign deployment, and performance reporting. The platform costs $3,600-$6,000 per year. An agency retainer costs $5,000-$15,000 per month.

Amazon's advertising revenue is projected to hit $94 billion by 2026: up 149% from 2022. Their self-service AI creative studio lets brands bypass agencies entirely.

Every major platform is building the same thing: a world where the advertiser talks directly to the AI, and the agency is not in the room.

The Holding Companies See It Too

The world's largest agency networks are not fighting this trend. They are restructuring around it.

WPP declared "we are no longer a holding company" and launched Elevate28: a plan to cut 9,000 jobs and save £500 million. Their CFO said the quiet part out loud: "In a business where most of our cost savings are people, that will mean a reduction of certain heads."

Omnicom doubled its cost-cutting target from $750 million to $1.5 billion. Killed three legendary agency brands: DDB (77 years old), FCB (153 years old), and MullenLowe. 4,000 direct layoffs plus 10,000 more impacted by sell-offs.

Dentsu put its $4.5 billion international business up for sale. Every buyer walked away. The stock crashed 11%.

The holding companies are not investing in the traditional agency model. They are dismantling it.

What the Data Shows

The numbers paint a clear picture of where the industry is heading.

Agency holding company market share has dropped from 44.6% of US ad spend in 2019 to 29.6% in Q1 2024. A 15-percentage-point collapse in five years. Meanwhile, worldwide ad spending grew 8.6% year-over-year in 2025.

The market is growing. Agencies are shrinking. The money is flowing directly from brands to platforms, skipping the agency entirely.

60% of US senior marketing leaders said they spend less on agencies as a direct result of AI capabilities. 82% of major advertisers now have in-house agencies, up from 42% in 2008. 91% of brands have moved advertising back in-house.

Forrester predicts 15% of agency jobs will vanish in 2026. After 8% cuts in 2025.

This is not a rough patch. This is a structural correction.

Where Agencies Still Win

The platforms can automate targeting, creative generation, and campaign optimisation. That much is clear. But there are capabilities they cannot replicate, at least not yet.

Cross-platform strategy. Google optimises for Google. Meta optimises for Meta. Amazon optimises for Amazon. None of them optimise across all three. An agency that can build and execute a unified strategy across platforms provides value no single platform can match.

Client-specific knowledge. Platforms operate on data. They do not know the client's internal politics, competitive positioning, or three-year growth plan. An agency that embeds in the client's business and builds strategy around context, not just data, is doing work AI cannot automate.

Creative judgment. AI can generate a thousand ad variations. It cannot tell you which one aligns with the brand's long-term positioning. It cannot navigate the tension between performance metrics and brand building. That judgment comes from experience and understanding, not algorithms.

Integration and systems. The agencies that thrive are the ones building automated workflows that connect what the platforms cannot: CRM data, offline events, email sequences, content calendars, client reporting. These are custom systems that sit above the platforms, not inside them.

The Uncomfortable Middle Ground

Here is the hard truth for mid-size agencies.

You are too small to restructure like WPP. You do not have $500 million to invest in AI transformation. But you are too big to ignore the shift. If your agency still charges clients to do what platforms now handle for free, targeting, creative testing, delivery optimisation, your client is going to ask the question you do not want to hear.

"Why are we paying you for this?"

The agencies that have a clean answer are the ones building systems on top of the platforms. Custom reporting. Automated content production. Integrated workflows their clients cannot replicate in-house.

The ones without an answer are the ones OpenAI's head of monetisation was talking about when he said advertisers should just prompt ChatGPT and go.

What to Build Now

If you run an agency, here is the diagnostic question: if every major platform fully automated every campaign your clients run, what would your clients still need you for?

If the answer is "nothing," that is your problem, and the timeline to solve it is shorter than you think.

If the answer is "strategy, integration, creative judgment, and systems that connect everything," then your job is to make that answer so obvious to your clients that the question never comes up.

The platforms will keep trying to replace agencies. OpenAI failed this time. Google, Meta, Salesforce, and Amazon will not stop. The only defence is building capabilities that make you indispensable: not because the platforms chose not to replace you, but because they cannot.

OpenAI told every advertiser to fire their agency. Their own ads failed. But the next attempt will not. Build your moat now, while you still have time.