We built an agency. Then we took 90% of it apart.

By Kevin Veitia, founder of OuiGrowth

Most agency problems aren't about bad people. They're about math. Here's the math, what it does to your account, and how we rebuilt OuiGrowth around it.

Headcount doesn't scale

An agency grows by adding clients. Every client needs hours, and hours need people, and people are the biggest cost on the books. So headcount always lags behind the client list, and the business settles into a model you've probably met.

A charismatic senior person sells you. Your account goes to a junior marketer, often fresh out of school, who is juggling 20 or so other accounts. Yours gets five to ten minutes of their day. The report on your monthly call was assembled in a scramble an hour before it. Nobody planned that out of malice. It's the business model working as designed.

Twenty accounts, a handful of slots

There's a hard limit under all of this, and it isn't effort. In 1956, the psychologist George Miller published the paper that gave us “the magical number seven, plus or minus two”: roughly how many things a person can hold in mind at once. Later research suggests the real number may be closer to four, which only strengthens the point. And those few slots aren't reserved for work. They're shared with everything else on your mind, from the next meeting to a child's dentist appointment.

David Allen built Getting Things Done, one of the most widely used productivity systems there is, on exactly this limit: your head is a poor place to keep a list. A manager carrying 20 accounts can't keep them all in mind. Most get the shallow version of attention, and no tool adds more slots.

Junk traffic, then the shrug

When attention is that thin, the easiest work wins. Link-click campaigns and boosted posts are quick to set up and make a dashboard look busy. The agency takes the fee, sends the traffic, and when nothing converts, the blame moves to your side: “We sent you 20,000 visitors. If they didn't buy, look at your website.”

What's usually missing is anyone looking behind that door. No conversion-rate work, no heat maps, no funnel analysis, nobody asking why visitors leave. And some of those 20,000 may never have been people. Here's how we found that out.

Of the agency-run accounts Kevin has audited over the years, he'd say exactly one was run by a team that knew what it was doing.

The remora and the shark

The second force is quieter. Agencies are a relationship business, and the most valuable relationship is with the CMO. When a new CMO arrives, the old agencies often go, and the new CMO brings in the people they already trust. The agency rides along like a remora on a shark.

And CMOs move on faster than almost anyone else in the C-suite. Spencer Stuart puts average CMO tenure at S&P 500 companies at 4.1 years. A 2025 study by Findem and CMO Huddles found a median of 36 months across US companies, and an average of 2.6 years at venture-backed ones. An agency that knows its champion has two or three years left has little reason to try anything ambitious. Hit the vanity metrics, keep the reports tidy, and ride the relationship until it ends.

Doesn't AI fix this?

It fixes the agency's margins, not your account. AI can make a junior faster, but it doesn't give them more attention, and it doesn't change the incentives. No tool makes an agency care about the years after its champion leaves. AI makes the broken model cheaper to run. It doesn't make it better.

What we did about it

We learned this from the inside. In 2024 and 2025, OuiGrowth grew to about ten people: account managers, videographers and more. Then contracts ended, the economics changed, and we had to decide what kind of business this was going to be.

When we looked honestly at the work, much of what we were paying juniors to produce came out of the same AI tools we could run ourselves, with an extra layer between the client and the person who actually understood the account. So we cut about 90% of the business.

What's left is small on purpose. Kevin leads every account and does the work himself, with AI doing the jobs that layer used to do. When a project needs a craft where AI still falls short, we bring in one of about three trusted specialists, mostly designers and videographers. AI-generated UGC is a good example: one usable video can take two hours and 50 credits, and half the attempts still miss. A good human creative is faster and better.

It's a smaller business than a bloated agency, by choice. In return, we only take accounts with a problem we find genuinely interesting, and we treat each one almost like our own business. That makes us loud and opinionated about it. It also means we don't stop at the ad. We look at the landing page, the funnel and the heat maps, because the job is your revenue, not our click-through rate.

Then why not just use AI yourself?

Fair question, and some founders should. But AI doesn't replace the expert. It replaces the layer of juniors around the expert. It's confident when it's wrong, it doesn't know which lever matters in your account this month, and it takes real time to steer. Someone still has to know what good looks like and notice when the output is junk.

That's the real shift. It isn't humans out and AI in. It's that the economics now favor lean, senior consulting over headcount-heavy agencies, because one experienced operator with AI can do what used to take a team. A bloated agency can't give you that, and neither can a founder alone with a chatbot.

Questions to ask any agency, including us

• Who will log into our ad account every day, and how many other accounts do they run?
• What does each campaign optimize for, and how will we know the results were people?
• Will you look past the click, at our landing pages and funnel?
• Who is accountable for results, and when did they last run an account like ours?

With us, the answer to the first and last questions is the same: Kevin.