By Federico Soto Roland — Strategy, AI & Digital Director, NSB Agency.

For roughly the last 12 months, my days have revolved around AI: it's my main focus. How to learn it, how to implement it in our agency, how to get ready for the seismic shift that is already here — and that is coming with far more force over the next few years.

Reports like AI 2027 — written in 2025 — got many of their predictions right, and by the end of this year models will be self-improving faster, powered by code generated by AI itself.

The pace of growth has accelerated incredibly. Just look at everything shipped in the last eight months: it's astonishing. What at the start of 2025 was crude and far off is today practically impossible to tell apart from the real thing.

There is no task in an advertising and brand communication agency that isn't being disrupted by AI: research, account planning, copywriting, image generation, video, commercials, web design, coding, media planning… everything is being turned upside down. But let's be clear about one thing: this is only the beginning.

That's why I wanted to ask Claude (my favorite AI) how it sees the business of agencies like NSB in 2030. The result is three scenarios: three possibilities with varying odds of occurring and very different effects depending on which one materializes.

Before getting into them, a piece of context that isn't a projection but the present: global ad spend is growing at record rates while the agency business is stagnant — the money flows to the platforms, not to the agencies. The holding companies cut tens of thousands of jobs and pulled century-old brands from the market. 82% of large U.S. advertisers already have an in-house agency. And Meta runs a system where you give it a URL and a budget, and it does the rest. That's the board on which the three futures are played.

What follows is CLAUDE's assessment.

Scenario 1: “Transformation, not extinction” — Estimated probability: ~55%

The base case, and the one I consider most likely. The agency function survives, but unrecognizable in its business shape.

AI reliably executes everything bounded and repeatable: variants, adaptations, versioning, operational media, directional research, reporting. But end-to-end creative-strategic flows with no human in the loop remain fragile for large brands — because the problem isn't generating content, it's being accountable for it.

The holding companies consolidate and slim down. Agile independents gain ground in niches. In-housing keeps growing, but brands still need to make outside what can't be “manufactured” inside: an external eye, cultural judgment, brand stewardship.

The viable agency in this scenario is smaller and more senior. The classic pyramid — many cheap juniors executing, a few seniors deciding — inverts: a few seniors with AI assistants.

The billable hour dies, not out of ideology but out of arithmetic: when a tool compresses a twenty-hour deliverable into five, charging by the hour means billing four times less for the same value. You charge for value and outcomes, or you don't survive.

Scenario 2: “AI expands the market” — Estimated probability: ~25%

The scenario almost no one discusses, and it's a shame, because it has historical precedent: every time a technology collapsed the cost of producing something, the market for that thing grew.

When producing quality creativity stops costing what it used to, thousands of mid-size and small brands — that could never afford professional strategy and creativity — enter the market as clients. The demand for judgment gets democratized. Agencies capture services that don't even exist as a category today: orchestration of AI systems, brand stewardship over content generated at scale, custom development of tools and agents.

And there's a regional chapter that hits close to home: in this scenario, the creative talent of markets like ours — proven on the global podiums, at a fraction of the U.S. cost, now supercharged by AI — competes on equal footing with anyone. Geography stops being destiny.

There are already thirty-person agencies winning national accounts away from networks with thousands of employees, taking projects from brief to launch in less than half the time. That case stops being an anecdote and becomes a pattern.

Scenario 3: “The platforms eat our lunch” — Estimated probability: ~20%

The uncomfortable scenario, and it would be dishonest not to put it on the table with its real probability, which is not negligible.

AI agents generalize faster than expected. Meta and Google reliably close the full advertising loop — from URL to result, with no intermediaries. Brands internalize strategy with their own data plus AI. The executable part of the business collapses, and agencies are reduced to high-level consulting for a handful of top brands: a much smaller, much more contested market, where only those already positioned on judgment and criteria — not execution — get in.

And this scenario has a silent aggravating factor: by automating junior work, the industry is demolishing the entry rung on which all of us who decide today were trained. If this future accelerates, the senior-talent crisis of 2035 is being manufactured right now.

Back to my own analysis.

What the three scenarios have in common

Here is, for me, the most important finding of the exercise: the three scenarios punish exactly the same thing and reward exactly the same thing. They punish execution as a business model and they reward judgment.

In any of the three futures, billing for producing volume is a losing position — only the speed of the defeat changes.

Perhaps the only defensible value we have as agencies lies in three things: creating and defending ideas with genuine cultural tension and high salience, in an ocean of content with a whiff of AI; brand stewardship when content is generated at industrial scale, and someone has to be the “no” button before publishing; and deep knowledge of specific brands, categories and cultures — which no generic model has (at least for now).

Regardless of which scenario materializes, the right strategy is the same in all three: move from execution toward judgment, from volume toward value, from hours toward outcomes.

The only thing that changes between scenarios is how much time you have to make that transition.

Based on my experience over these last 12 months with AI — doing, researching, learning and watching it evolve — I think we have less time than the industry believes.

AI isn't coming to kill the agencies. It's coming to do something more demanding: strip away everything that made us profitable without forcing us to be very good, and leave standing only what made us valuable.

The question, then, isn't where AI is taking us; it's which of the things we do today will keep us relevant five or ten years from now.