Why traditional agencies are dying — and how AI saved my 19-year career

In 2006 I founded a design agency in Chicago. For roughly ten years we held the #1 spot on Google for "Chicago web design" and ranked Top 3 globally for "web design." We worked with Subway, Unilever, Northwestern University, Tribune Media, and the City of Chicago. By every measure of the era, the model worked.

That model is now dead. Not wounded, not evolving, not "facing headwinds" — dead. And after nineteen years of building through every major shift in digital marketing, I'm convinced that's a good thing. But only if you understand exactly what died, because most of the industry has diagnosed it wrong. The obituary everyone is writing says "AI killed the agency." That's not what happened. AI killed one specific part of the agency — and in doing so, it exposed which agencies ever had the other parts at all.

What the old model actually sold

Strip away the case studies and the awards shelf, and a traditional agency sold three things bundled into one retainer:

• Labor — hours of production: design, copy, edits, resizing, publishing, reporting.
• Judgment — knowing what's worth making, for whom, and why it will work.
• Access — specialist skills a client couldn't justify hiring full-time.

The bundle was the business model. Clients couldn't buy judgment by itself, so they bought the hours it was wrapped in. Agencies priced the package on labor because labor was countable — and for two decades, that arithmetic held. When I ran Blueprint, a campaign genuinely took a team weeks. The invoice and the effort matched, and nobody had a reason to look too closely at which third of the bundle was actually creating the value.

Tools generate content. Judgment decides what's worth generating — and owns whether it works.

The retainer math that stopped working

Here's the version of this story nobody puts in a conference talk. A mid-size retainer used to buy, say, 120 hours of monthly production: concepts, drafts, revisions, adaptations, scheduling, a report. The client was really paying for the ten hours of judgment inside those 120 — the positioning call, the creative direction, the "no, not that" moments. The other 110 hours were the delivery vehicle.

Generative AI collapsed those 110 hours into something closer to eleven. Not for every task, not perfectly, and not without supervision — but close enough that any client who has watched someone competent use these tools can no longer un-see it. When the labor evaporates, a labor-priced invoice becomes indefensible. The agencies in real trouble right now aren't losing a technology race. They're discovering, in front of their clients, that labor was the only layer of the bundle they ever reliably delivered.

What AI actually replaced

The production layer — the hours of execution that used to justify retainers. In practice, that means:

• First drafts, variations, and resizing — gone from billable to nearly instant
• Research and competitive teardowns — compressed from days to hours
• Repurposing — one strong idea becomes a video script, posts, graphics, and an article in a single working session
• Publishing, distribution, and first-pass reporting — automated end to end

I felt this personally before I ever consulted on it. By the time generative AI arrived, I'd already spent years automating content pipelines with Make, n8n, and custom integrations for my own projects. When the new tools landed, I recognized them immediately: this wasn't a gimmick. It was the production department, as software. The work I once staffed a studio for — thousands of websites, campaigns, and digital experiences over a decade — now needed a system and a fraction of the people.

What AI made more valuable

Taste, strategy, and accountability. When everyone can produce infinite content, knowing what's worth producing becomes the entire business.

This is the part the doom narrative misses. Every previous shift I built through rewarded exactly the same thing. When search arrived, ranking tricks were commoditized within a few years — sustained strategy is what kept us at #1 for roughly a decade. When social arrived, posting was free for everyone — judgment about format, hook, and timing is what drove the releases I marketed to #1 on iTunes Thailand ahead of global artists, and grew pages past a million followers. Now production itself is commoditized, and the premium moves — again — to the layer above it.

Scarcity didn't disappear. It relocated. It used to live in production capacity. Now it lives in three places: knowing what's worth making, making it distinctively, and being accountable for whether it worked. None of those can be generated.

The new agency is a system, not a staff

The operators winning right now redesigned around this reality instead of resisting it. The structure that replaces the old bundle has three parts:

A strategist at the top

A human decides the positioning, the audience, the offer, and the standard. This is the judgment layer clients were always really paying for — now sold openly instead of hidden inside hours.

A production system underneath

AI-assisted pipelines handle research, drafting, variations, and distribution — with human taste applied at the two points where AI is weakest: the hook and the edit. The output is the same as the old agency's — attention, authority, leads — but the cost structure and the speed are unrecognizable.

Accountability at the end

Because the system is instrumented, results are visible. Effort gets reallocated to what the data says is working. "We were busy this month" stops being a deliverable.

This is precisely how I rebuilt my own career. The studio era gave me the judgment; the automation era gave me the system. Today that structure runs as Design Delulu — managed content growth systems built and operated for businesses — and it's why I'd argue AI didn't end my nineteen years in this industry. It compounded them.

If you're hiring — or running — an agency in 2026

The practical version of everything above is a short list of questions. Ask them of any agency you're evaluating, or of your own shop:

• What am I paying for: hours or outcomes? If the answer is a deliverables count, you're buying the layer AI already ate.
• Where does AI sit in your process? "We don't use it" and "we use it for everything" are both wrong answers. The right one names exactly where human judgment stays in the loop.
• Who decides what's worth making — and on what evidence? Strategy should be visible, not implied.
• What happens when something underperforms? A system has a feedback loop. A vendor has excuses.
• Could your team run it without them? The best modern operators build systems that survive the handover — and say so up front.

The takeaway

The mental model to keep: every technology shift commoditizes execution and re-prices judgment. Web design, search, social, and now AI — the pattern hasn't broken once in twenty years. The agency model isn't dead; the labor-priced version of it is. What replaces it is a content system with a strategist at the top, automation underneath, and accountability at the end. That's what I build today, and it's the only model I can honestly recommend to a client with a straight face.

Frequently asked questions