Why AI has made the agency pitch worse

Sometime in the last two years, a new slide has quietly crept into every agency’s credentials deck. It sits somewhere after the case studies and before the team page. It features a list of logos – Midjourney, ChatGPT, Claude, Firefly, sometimes a few more obscure ones – arranged to signal technological fluency. The headline above them says something like: “AI-powered creativity” or “The future of brand” or, most ambitiously, “Human intelligence, amplified.”

The slide is, in almost every case, meaningless. And its widespread adoption has made the agency evaluation process measurably worse for the clients trying to navigate it.

Here is the problem. When every agency claims AI capability, AI capability stops being a differentiator and becomes a hygiene factor – a box that must be ticked to stay in the pitch. The claim is rarely substantiated. The logos on the slide tell you which tools the agency has subscriptions to. They say nothing about how those tools are used, by whom, in which phases of work, and – most importantly – what they actually change about the quality or speed of what gets delivered. A junior account team with a ChatGPT subscription and a senior principal who uses Claude to run research synthesis in two days instead of three weeks are both AI-enabled. They are not the same thing.

The consequence for clients is a new and genuinely difficult evaluation problem. In the pre-AI era, assessing an agency’s capability was imperfect but tractable. You looked at the work. You met the team. You probed the process. You made a judgement about whether the people in the room would still be in the room when the work was being made.

Now there is an additional variable that most procurement processes are not equipped to evaluate: whether the AI use an agency describes is real, meaningful, and appropriately governed – or decorative.

Decorative AI use looks like this: the agency uses AI to generate options quickly in pitches, presenting more work in less time to win the business, then reverts to standard workflows in delivery. Or it uses AI for production tasks – image generation, copy variants – without any corresponding change to the strategic quality of the work. Or it describes AI as part of the process without disclosing which specific decisions AI touches and which remain human.

None of this is necessarily dishonest. But it produces a pitch that overstates the operational reality of how the work will be made. The client buys AI-accelerated strategy and receives standard-pace execution with a Midjourney image in the deck.

There is a framework for cutting through this, and it involves four questions that almost no pitch evaluation process currently asks.

1. Which specific phases of your process does AI accelerate, and by how much?
   A serious answer to this question is granular. Research synthesis, three weeks compressed to four days. Positioning territory exploration, four options explored simultaneously instead of one at a time. Copy drafting, first drafts in hours rather than days, with human editorial layer applied before client review. A vague answer – “AI helps us move faster and explore more” – is not an answer. It’s a brochure.
   ‍
2. Which decisions in your process remain exclusively human, and why?
   This question reveals whether the agency has thought seriously about where AI judgement ends and human judgement must take over. Strategic positioning decisions. Ethical assessments of content. Judgements about cultural sensitivity or brand risk. Client appropriateness. An agency that can’t answer this question with specificity either hasn’t thought about it, or is using AI in places where they probably shouldn’t be.
   ‍
3. How do you disclose AI use to clients, and what governance sits around it?
   Do they tell you which tools were used in which phase of your engagement? Do they have a data-handling policy for client information that touches AI systems? Do they know which of their AI platforms train on submitted content by default? If the answer to any of these is uncertain or vague, the AI use is not governed. Ungoverned AI use is a reputational and commercial risk that the client, not the agency, will wear if something goes wrong.
   ‍
4. Can you show me work where AI changed the strategic outcome, not just the production speed?
   This is the hardest question and the most revealing. Production speed is real and valuable but relatively easy to achieve. Strategic impact – a research phase that surfaced an insight that changed the campaign platform, a competitive analysis that identified a positioning territory the client had never considered – requires a level of AI integration that goes beyond tools and into methodology. If the case study answers are all about faster timelines rather than better thinking, the AI use is production-level, not strategy-level.

None of this is an argument against AI in agency work. The opposite is true. Agencies that have genuinely integrated AI into their strategic process – not just their production workflow – are delivering better outcomes faster than those that haven’t. The research depth, the iteration speed, the breadth of strategic territory that can be explored in a fixed timeframe: all of these improve materially with serious AI integration.

But “serious AI integration” and “has an AI slide in the credentials deck” are not the same category of thing. The clients who will get the most from the current agency landscape are the ones who learn to tell the difference – before the pitch, not after the campaign.

The four questions above are a start. Use them. The agencies that answer well deserve the work. The ones that don’t will tell you something important about how they’ll handle the next brief too.

‍