The AI Output Illusion
AI is about to expose the biggest lie in the agency model.
A friend of mine is paying his Meta agency $25,000 a month. Ads, creatives, landing pages, campaign management. The full stack.
Last week someone in our group chat posted a workflow built by Codex — research engine, strategy briefs, creative generation, distribution, analytics loop. The whole thing. Built in an afternoon.
His response: “I think this entire model is about to change.”
He’s right. But probably not in the way most people think.
Right now, the AI conversation is almost entirely about output. New tools, new workflows, new prompts. Someone generates a landing page in 30 seconds and posts a screen recording. Someone else builds an entire app from a voice note. The demos are impressive.
But there’s a question almost nobody is asking: output of what, and why?
Because generating something is not the same as knowing what should exist.
AI made output cheap. You can generate ten ad variants, ten landing pages, ten positioning angles in minutes. Soon it’ll be hundreds.
But cheap output without direction is just noise at scale.
And the noise is creating something else: FOMO. People feel like they’re behind. Like something important is passing them by. So they burn through tokens generating things they’ll never use — apps they don’t need, content with no audience, automations with no purpose. Activity that looks like progress but produces nothing.
For years, companies paid agencies for the operational layer — campaigns, copy, graphics, media buying. That work was expensive because it required time and people.
I know this because I ran an agency. And I can tell you: trying to charge a client for strategy was almost impossible. Nobody wanted to pay for thinking. They wanted deliverables. Tangible output. Strategy was treated as overhead — something that should come free, bundled with the “real work.”
So I learned to hide it. I embedded my strategy into the outputs themselves, and that became my edge. The client thought they were paying for execution. They were actually paying for decisions — they just didn’t know it.
AI is about to make that dynamic visible.
When output becomes cheap, fast, and abundant, the difference between someone who generates and someone who knows what to generate becomes obvious. The strategic layer can no longer hide inside the operational cost. It has to stand on its own.
Most people right now are in the tools phase. Experimenting with prompts, testing models, building automations. That phase is natural — it happened with every new technology.
But tools are not the game.
One model generates ads. Another analyzes data. Another writes content. Another builds code.
Someone still has to decide: what problem are we actually solving? What data matters? What signals do we follow? What system learns from the results and adjusts?
Without that layer, AI is a very fast machine for producing things nobody needs.
The common fear is that AI will replace people. But what’s actually happening is a reorganization. The value isn’t disappearing — it’s shifting.
The strategists, the architects, the people who always knew what should be built but were bottlenecked by the cost of building it — they now have tools that let them execute at speed. That’s not a loss. That’s leverage.
What is losing value is pure execution without direction. The person who only knew how to produce deliverables but never decided what to produce or why. That role was always dependent on someone upstream making the right calls. Now the upstream person doesn’t need a team of twenty to act on those calls.
This is a restructuring of companies, teams, offers, and business models. Not because tools are powerful — tools have always been powerful. But because for the first time, the tool layer is cheap enough that it stops being the bottleneck.
The bottleneck is now clarity. Direction. Knowing what should exist and why.
And that was always the hard part. We just couldn’t see it behind the cost of everything else.
Lucas Hubert