The agency business model is dead. Or is it?

Every six months someone announces the death of the agency business. I've been reading the same post since 2019. It's always wrong, and it's always a little bit right. The agency business model isn't dying. The agency business model is bein

Every six months someone announces the death of the agency business. I've been reading the same post since 2019. It's always wrong, and it's always a little bit right. The agency business model isn't dying. The agency business model is being replaced by a smaller, sharper thing that still calls itself an agency because it hasn't found a better word. I've run one for five years. Here's what I actually see from inside.

What's dying

The ten-person retainer shop. The "we do everything" pitch deck. The hourly rate card. These are dying faster than most people in them realize. I've watched two friends close agencies this year that had real revenue and real clients six months before they closed. Neither of them was a bad operator. Both of them were running the wrong shape.

The client who wants to pay for hours is rare. The client who wants to pay for a result is common. An hours-based agency cannot survive long next to a product-priced competitor, because every pitch becomes a conversation about the rate, and every delivery becomes a conversation about the timesheet.

I don't think AI killed them. I think AI was the thing that exposed they were already dead and coasting.

What's surviving

Small, obsessive, productized. Four people or fewer. One offer, one price, one deliverable. The agencies in my circle that grew this year all look like this. The ones that died looked like generalists.

The offer shape matters. "We build AI chatbots" is an agency. "We build a customer support agent for Shopify stores that cuts first-response time to under ninety seconds, and it costs $14,000" is a product with a service wrapped around it. The second one is what clients are buying. The first one is what's on the generic pitch deck.

I priced my first productized offer in early 2024 and lost money on it. I priced my fifth one six months later and stopped taking hourly work. The delta between those two prices was not pricing skill. It was learning, in public, what the deliverable actually costs to make.

The "AI agency" trap

Everyone on LinkedIn announced an AI agency this year. The positioning is so crowded it functions as anti-positioning. If you say "AI agency" in a cold email, you're in a pile of a thousand nearly identical messages. The word does no work for you.

The agencies winning the AI wave are not calling themselves AI agencies. They're calling themselves what the client already calls their problem. "We automate revenue operations for Series B SaaS." "We deploy agents inside finance teams." "We ship multilingual content engines for DTC brands." AI is the how, not the what. The positioning is the vertical or the function.

I made this mistake for about eight months. Called us an AI agency. The cold-email reply rate was under one percent. We changed the line to name a specific outcome for a specific buyer, and the rate went to four. Four percent is not a miracle — it's just what happens when you stop sounding like the pile.

The product trap, in reverse

There's a counter-trap. Agency operators get told to "build a product" and half of them do. Most of those products die. A services business doesn't easily become a product business, because the DNA is different. Services DNA says yes to custom. Product DNA says no.

The middle path I've watched work: the agency stays an agency, but it builds internal products that show up inside the deliverable. An onboarding tool used on every engagement. A dashboard the client gets as part of the offer. A document generator that cuts delivery time by half. These aren't SaaS — they're leverage. They turn the agency into a smaller team doing the same work.

I know three operators who followed this path in 2025. Each of them runs a four-person team doing more revenue than their former ten-person version. They didn't become product companies. They became agencies with product economics on one axis.

I stopped trying to build a product and started using what I would have built. Suddenly my agency margins looked like a product's. Turns out I needed leverage, not a second business.

— an old operator friend, after his third failed SaaS

The price conversation

Here's the part nobody writes. If your agency still sends quotes by hour, you are not in the same market as the agency next to you that sends quotes by deliverable. The hourly-billed agency is in a rate market. The productized agency is in a value market. These are two different games with two different economics.

The rate market is a commodity market. It ends with a race to the bottom you can't win against someone with a lower cost base. The value market is a wedge market. It ends with the client saying yes to the number or no, with no middle. I know which one I'd rather be in.

Moving a book of business from rate-based to value-based is slow. It took me eighteen months. The first offer was wrong. The second was right for the wrong buyer. The third one worked, and after that each new offer took about a month to design because I was doing it in the shape I had learned.

The freelancer question

I get asked whether small agencies will be replaced by well-equipped freelancers. Partially. A senior freelancer with Claude Code and a tight niche can do things a ten-person agency used to need to do. Clients know this. Some of them prefer it.

But there's a piece freelancers rarely have and a small agency can: the continuity layer. Someone answering when the usual person is out. A second pair of eyes on a delivery. A junior to push work to. Two operators who trust each other outperform one brilliant freelancer for most kinds of ongoing engagement, and the math is not close.

If you're a freelancer asking whether to hire one person, the answer is usually yes. It's the smallest step toward the shape that works, and the one you will wish you'd taken a year sooner.

So, is it dead

No. Not the form that's earning. The category is fine. The specific version most of us learned — generalist, ten FTEs, hourly billing, custom everything — is dying at a clip that is hard to watch.

What comes next looks more like a studio than an agency. Three to five people. One productized offer that pays the rent. Bespoke work that pays for the summer. A brand that is distinct enough to be named in a conversation among peers. Software where it helps. No software where it doesn't.

I called it an agency for five years. I'm thinking about calling the next version a studio, not because the label matters, but because the word "agency" is doing less and less to explain what I actually sell. The word will follow the work.

If you're running one: look hard at your offer shape before you blame the market. If you're starting one: pick a narrower wedge than feels comfortable. If you're leaving one: the industry is not the problem you think it is.