The Pitch Collapse: When AI Starts Evaluating Itself

Satire, certainly. But satire that hits home because everyone in the industry recognises the reality behind it.
In his announcement, Sadoun was more explicit: the combination of overpromising on AI and making commercially unsustainable offers in pitches was driving “massive job cuts” across the industry.
His call to action was clear: “Collectively, we have to stop this race to the bottom and reaffirm our unique ability to deliver what clients really want.”

Sadoun describes the loud, commercial side of the problem: agencies are competing on promises of efficiency rather than the strength of their ideas.
But something quieter, and potentially more consequential, is happening within the pitch process itself. Nowhere is this more apparent than in procurement processes governed by formal rules, particularly in the public sector.

Marketing departments increasingly use AI to write their agency briefs.
Agencies then use AI to develop strategies and creative routes based on those briefs.
Before submitting their proposals, agencies turn to AI once again to check how closely their work aligns with the brief. And on the other side of the table, another AI system assesses those submissions against the client's evaluation criteria.

Four rounds of AI, with very little human friction in between. The result? A great deal of mediocrity and an increasing tendency towards sameness. AI research has a name for a related phenomenon: model collapse.
When models are repeatedly trained on their own synthetically generated outputs, the edges of the distribution begin to erode. The rare, the unconventional and the unexpected gradually disappear until little remains beyond the statistical average.
Each generation becomes slightly more like the one before it, without adding anything genuinely new.

Something similar is now happening in agency procurement. Except it is not the models that are collapsing. It is the quality of the pitches.
Pitch collapse is the downward spiral that occurs when both sides of a procurement process use the same technology to produce and evaluate exactly what the other side, using that same technology, already expects.

Inside an agency, it might sound like this:
“We're significantly more efficient now. We can respond to ten tenders at once.”
Unfortunately, that rarely leads to an idea that genuinely moves a brand forward.

Orlando Wood would hardly be surprised.

His assessment of the advertising industry is that, over several decades, it has systematically favoured salesmanship, which is literal, linear and focused on rational proof, over showmanship, which embraces lateral thinking and the power of surprise.And this shift has had measurable consequences for advertising effectiveness.

AI amplifies this imbalance rather than correcting it.
A large language model is, structurally speaking, a salesmanship machine. It recombines what has already been said according to what is statistically most likely to persuade. This aligns with what creativity researcher Margaret Boden calls combinational creativity: bringing existing elements together in new configurations. What is missing are the exploratory and transformational dimensions of creativity, particularly the ability to challenge the underlying system itself.

To be clear, there are plenty of stages in agency procurement, particularly in public-sector tenders, where automation makes perfect sense. Pitches still cost agencies countless unpaid hours, often translating into mid-to-high five-figure or even six-figure sums. A significant proportion of that effort goes into completing forms, providing supporting documentation, compiling case studies and assembling team CVs. Using AI to organise tender documents, systematically complete pre-qualification submissions or cross-check evaluation matrices can save valuable time and resources. Time that can then be invested in the work that actually matters.

The problem begins when providers apply the same efficiency logic to the creative pitch itself. There are already AI-powered solutions openly advertising the ability to complete a pitch in just three hours. Whether that figure is accurate is almost beside the point. What matters is what is being sold: the elimination of precisely that part of the pitch process where the hard work is not a flaw in the system to be fixed, but the price of arriving at an idea worth having.

Writing in The Media Leader, consultant Paul Evans captures the underlying dynamic: years of efficiency-driven optimisation have made the major agency holding networks increasingly indistinguishable. AI now makes that interchangeability costly by exposing where genuine differentiation has disappeared. That is why pitch collapse disproportionately affects the middle of the industry. Agencies that already sounded almost exactly like everyone else now sound exactly like everyone else.

The trap lies in believing that AI can deliver the idea. Every agency on the shortlist is using the same AI assistants.
Chances are the client has already used those tools to explore their own market. A pitch built entirely on AI-generated research simply tells the client what they already know. What wins is the observation that cannot be found in any dataset. A revealing comment from a conversation with real people in the target audience.
A supermarket shelf that tells a different story from the sales figures. An anomaly in the client's own data that nobody has noticed before.

AI's real value lies in clearing away enough of the surrounding work to give us time to uncover precisely those insights. AI can also play a valuable role in the pitch process as an opponent rather than a supplier. As a challenger that interrogates every idea, asks the uncomfortable questions and puts every recommendation through a rigorous stress test.

But what emerges from that back-and-forth remains a human responsibility, including the judgement to reject AI-generated suggestions when they don't hold up. The advertising industry's longstanding problem with creative courage has simply acquired a new, more convenient excuse: artificial intelligence.
Anyone who uses AI in a pitch to avoid thinking, rather than to create more time for thinking, has already lost.
Before the competition has even begun.