A great pitch is no longer proof of a great agency
AI has dramatically lowered the cost of producing convincing work. That changes what founders and CMOs should actually test when choosing an agency or growth partner.
For years, marketing pitches were built around a fairly simple assumption.
If an agency could produce a strong strategy, a compelling deck, sharp creative ideas and a convincing presentation, there was a reasonable chance it could deliver strong work after winning the account.
Within days, a team can create market analysis, segmentation, strategic hypotheses, communication platforms, visual concepts, product mock-ups, prototypes, data models and even functioning AI agents.
The problem is not that these outputs are fake.
Many of them may be genuinely good.
The problem is that they tell us less than they used to about what the organisation behind them will be able to deliver consistently over the next twelve or twenty-four months.
AI has made it easier to demonstrate a capability than to actually possess it.
And that changes the logic of agency selection.
When a good pitch was expensive
Around a decade ago, I remember listening to a panel of CEOs from some of the leading advertising agencies discussing the economics of pitching.
The numbers being mentioned for preparing a single major proposal ranged from the equivalent of a few thousand euros to well into five figures.
Agencies were frustrated by the amount of senior strategy, creative, research and management time being consumed by competitive pitches.
There was even an industry push for pitch fees of around €1,000.
That amount did not come close to covering the true cost of preparing the proposal.
But it did something else.
It created at least a small economic signal that the advertiser was serious about the process, that the number of invited agencies might be rationalised and that the pitch itself had a real cost.
The cost of producing a sophisticated response acted as a filter.
An agency had to decide whether it was worth mobilising strategists, creatives, researchers, designers and senior management.
Today, part of that cost has collapsed.
Research can be accelerated.
Initial strategic hypotheses can be generated in minutes.
Dozens of communication directions can be explored almost instantly.
Visual concepts can look production-ready.
A prototype can resemble a mature product.
A technology demonstration can appear close to implementation.
A presentation can be beautifully structured and visually polished.
This does not mean agency work has become easy.
It means something more subtle.
The quality of the pitch can now improve much faster than the underlying capability of the organisation delivering it.
The pitch can look better than the agency
A few years ago, producing an exceptionally sophisticated pitch required capabilities that at least partially reflected the capabilities of the organisation itself.
That relationship is becoming weaker.
A small, highly capable pitch team supported by AI can now produce an extraordinary amount of high-quality material.
It may create deep research, dozens of creative directions, professional visualisations, a product prototype, a working technology demo and sophisticated data analysis.
All of that may be excellent.
But the client now needs to ask an additional question:
Am I looking at the capability of the organisation, or the capability of the pitch team to create an exceptional demonstration?
Those are not necessarily the same thing.
Buying a brilliant pitch without examining the organisation behind it is a little like buying an architectural render and assuming you have bought the building.
The render may be beautiful.
It tells you much less about the foundations, the engineering, the contractors, the operating systems or whether the structure will work five years from now.
An agency can win with a capability it does not yet have
This may be one of the more interesting consequences of AI for contemporary pitches.
Imagine an agency presenting an AI system that analyses thousands of customer comments and turns them into a live map of consumer insights.
The system works.
The demo is impressive.
The proposed application makes strategic sense.
But what exactly is the advertiser looking at? An established agency capability? A prototype created specifically for the pitch? A solution assembled by one unusually talented individual? A third-party technology? Or simply a convincing demonstration of something the agency believes it could build after winning the account?
None of these situations necessarily involves deception.
The agency may genuinely be able to deliver the solution.
But the buyer needs to know what is already real and what is still a promise.
The distinction between capability and prototype has become much thinner.
Which means the old question:
Can you show us that you can do this?
is becoming less useful.
A better question is:
Can you show us that you can do this repeatedly, with our data, inside our organisation, with the team that will actually work on the account?
That is a much harder test. And a much more valuable one.
Your brief may reach ChatGPT before it reaches the strategist
There is another consequence of AI that is less visible but equally important.
An advertiser sends a brief.
Inside it may be sales results, margins, product plans, customer research, budgets, competitive information, internal problems and expansion plans.
Fifteen minutes later, some of that information may already be inside an external AI tool.
Not because somebody intends to breach confidentiality.
Because this is increasingly how modern work gets done.
A strategist wants a synthesis of several documents.
A planner wants to identify patterns.
A copywriter wants to interrogate customer insights.
A designer wants to explore initial visual directions.
AI is becoming part of the operating system of agency work.
The obvious response is to ban it.
Add a clause to the NDA prohibiting external AI models.
But is that really what the client wants?
A complete ban may protect information while simultaneously preventing the agency from using tools that could significantly improve speed, breadth of analysis and productivity.
The more useful question is not: Can the agency use AI?
It is: How can the agency use AI with our information?
That conversation should increasingly include which tools are permitted, which information may enter them, which categories of data are prohibited, how outputs are verified, where data is stored, what happens to working materials and who remains accountable when AI-generated work is wrong.
A few years ago, these questions barely existed in agency selection processes.
Today, they are becoming part of basic due diligence.
The problem is not that agencies use AI
Quite the opposite. An agency that refuses to use AI in 2026 may simply be slower, more expensive or less capable than competitors using it intelligently.
The problem is that many advertisers still select agencies using signals designed for a pre-AI world.
They compare presentations.
Case studies.
Creative concepts.
Teams.
Timelines.
Fees.
All of these still matter.
But they no longer tell the whole story.
The more important question is becoming: What value does this organisation add beyond the technology available to everyone else?
That value may come from category experience.
From understanding the commercial reality behind the brief.
From identifying the wrong problem before spending money solving it.
From rejecting plausible but weak ideas.
From prioritising when there are twenty viable directions.
From understanding implementation constraints.
From recognising risk.
From knowing what has worked before and, equally importantly, why it failed.
From being willing to take responsibility for a recommendation.
From having people capable of changing course when reality stops matching the deck.
AI can expand the number of possible answers almost infinitely.
The agency's value increasingly lies in knowing which answers deserve to survive contact with reality.
Perhaps we should reverse the pitch
AI also exposes an older weakness in the way many companies buy marketing services.
A company creates a detailed brief, sends it to several agencies and compares their responses.
By doing so, it has already made one enormously important decision:
It has defined the problem.
But what if the diagnosis is wrong?
The client believes it has a communication problem.
One agency sees a product problem.
Another sees a distribution problem.
A third believes retention is the real constraint.
A fourth thinks the positioning itself is broken.
The traditional pitch often rewards the agency that provides the best answer to the question it was given.
Not the agency that recognises that the question itself may be wrong.
This is where a reverse brief becomes interesting.
Before asking agencies to produce complete strategies and detailed proposals, invite a smaller number of teams to challenge the diagnosis.
Do not start with:
How would you solve our problem?
Start with:
What do you believe our real problem is?
Only then define the scope of the actual pitch.
This is not merely a procurement technique.
It reflects a much broader principle of growth strategy.
A company can run a flawless selection process, choose an excellent partner and execute brilliantly against the wrong diagnosis.
The process worked.
The decision architecture did not.
AI makes this distinction more important, not less.
When good answers become cheaper to produce, the quality of the question becomes more valuable.
So how should agencies be evaluated now?
The answer is not to abandon pitching.
A pitch remains useful.
But it should increasingly be treated as the beginning of verification rather than the final proof of capability.
Advertisers can ask agencies to show which hypotheses they considered and rejected, distinguish what already exists from what still needs to be built, explain where AI was used, demonstrate a comparable process from a real engagement, introduce the people who will actually run the account, show a case where an initial recommendation turned out to be wrong, or solve part of the problem together with the client rather than only presenting a finished answer.
The objective is not to catch agencies using AI.
It is to understand where the enduring capability actually sits.
Is it in the technology?
In a few senior individuals?
In a repeatable organisational process?
In category expertise?
In proprietary data?
In decision-making quality?
In execution?
In access to talent?
In the ability to integrate all of those things?
Those distinctions are becoming more valuable than another fifty slides.
A great pitch can still mean a great agency. It just no longer proves it.
AI has not made good agencies less valuable.
It may do the opposite.
The strongest organisations can use AI to analyse more, prototype faster, explore more possibilities and increase the productivity of exceptional people.
What AI has changed is the reliability of some of the signals buyers have traditionally used to identify those organisations.
Research can be accelerated.
A deck can be improved.
Visuals can be generated.
A prototype can be built.
Dozens of strategic directions can be explored almost instantly.
So the ability to produce an impressive pitch is becoming less scarce.
The real question for founders, CMOs and procurement teams is therefore shifting.
Not: How much can this agency show us before we hire them?
But: What does this organisation contribute that will still matter after the pitch is over?
That is the capability worth buying.
Because AI has made a great pitch much easier to produce.
A great agency is still much harder to build.
