Part V
AI for creative agencies: thinking tools and execution tools
Most early demonstrations of generative AI were wonderfully easy to understand. Here is a picture of a car. Make it blue. Here is a description of an image. Create it. Here is a headline. Give me another hundred.
The input was visible, the output was visible and the improvement in speed was ridiculous. It is hardly surprising that much of the agency conversation about AI concentrated on execution: images, copy, video, adaptations, variations and production. That is where the technology first made itself obvious. It has also distorted how agencies think about where AI belongs.
Deciding what deserves to be made and producing it
Thinking does not end when execution begins. A copywriter writing a script is thinking. An art director finding the visual expression is thinking. A director can transform or discover the idea through the way it is shot.
The distinction I find useful is between upstream decisions about what deserves to be made and downstream production, expression and variation. Once the broad idea and intended change are established, the work moves into scripts, assets, formats and channel adaptations. Those activities still require judgement and can send the team back upstream when the idea changes.
AI is useful across both sides. The imbalance is that agencies applied it to production first while spending much less time on the decisions that determine which work should exist at all: the problem, audience, brand role, intended change and direction.
The spark has a history
Strategy has a mythology rather like creativity. We like the strategist who goes away, thinks deeply and suddenly has the answer in the shower, on a train or halfway through a conversation. Those moments happen. We just tend to forget everything that came before them.
The strategist read the brief, noticed contradictions, looked at the category, absorbed the research, understood the competitors and remembered something from another category five years ago. They probably rejected a few obvious answers without writing them down. Then something clicked. What we see is the click, while most of the process remains invisible.
Sometimes that process is painfully explicit. Someone fills in a framework. Consumer. Competition. Company. Boxes get completed and slides get made. Sometimes an experienced strategist barely notices they are doing it. The process is still there.
Some of the upstream work can be encoded
I know that sentence will irritate some strategists because it can sound like a claim that strategy can be automated. That is a larger claim than I am making.
You can encode the fact that a client brief should be interrogated before it is accepted. You can encode criteria for useful audience work and require a positioning to make a choice. This does not automate the entire strategic act. It makes upstream tasks available to the same technology agencies already use downstream.
None of this guarantees a brilliant strategy, but it makes it less likely that somebody skips half the thinking and jumps straight to an answer that sounds strategic. Experienced strategists already carry versions of these processes in their heads. Encoding them makes some of that invisible machinery visible.
Execution is easier to demonstrate
Making fifty adaptations used to take three days and now takes an hour. There is your business case. Strategic acceleration is harder to show. What is the value of noticing that the brief is solving the wrong problem? How much was saved because somebody challenged an audience assumption before the creative team spent two weeks working from it? What is the return on giving a strategist time to recognise a pattern instead of assembling the slides containing it?
These things can have enormous consequences and are difficult to put into a before-and-after demo. Creative agencies have always had this problem. Producing something is visible. Thinking that prevented the wrong thing from being produced usually is not. AI has simply inherited that bias.
Execution abundance changes what is scarce
Thinking and execution tools both release capacity. Agencies can use it to reduce cost, explore another direction, give smaller clients more attention or let senior people concentrate on decisions where experience matters. Cost reduction is sometimes necessary. As an entire AI strategy, it offers a temporary advantage when competitors can buy similar technology.
AI creates supply inflation. When every team can generate a hundred ideas, assets and versions, producing another option becomes less scarce. Selection, coherence, direction and taste become more valuable because somebody still has to decide what deserves to survive.
Agencies have missed this by applying AI downstream first. The next useful investment sits where problems are framed, options are rejected and the work acquires a direction before anybody asks the model to make the car blue.