What I'm Actually Building
It started with efficiency. It ended up being about soul.
Three weeks ago, Madison Utendahl closed her agency.
Not because she failed. Utendahl Creative had won every award worth winning, had millions in revenue, and was doing work that most people in the industry would kill to have on their reel. She closed it because the model is broken. Lower fees to keep the team, which means more clients needed, which means more pitching, which means burnout. Hiring full-time for three-month needs, hoping work would follow. Fixed costs in a market that was repricing everything underneath her.
"You're just still moving," she wrote about agencies that haven't adapted.
I keep coming back to that line. Because I manage two teams at the same time, and some days I'm not sure which one is still moving and which one is actually going somewhere.
D&L is a 14-person studio. We produce work for major brands. We're good at what we do. We've adopted AI tools and we produce faster, iterate quicker, handle more volume than two years ago. And if you asked me what we're selling, I'd give you an answer. But I'd be aware, as I gave it, that the answer has gotten blurrier. Not because we're worse. Because the things that used to differentiate us have become table stakes.
JackandAI is five people. Built with AI from the start. Different structure, different model. But the real difference is not structural. It's about what we're optimizing for.
Most of the industry is using AI to do more of what they already do, faster. More content. Lower cost. Higher volume. The efficiency case is real. But I've watched that logic play out close enough to see where it leads. You optimize for the metric and lose the thing the metric was supposed to measure.
The margin revolution that misses the point
Y Combinator put AI-native agencies on their Spring 2026 "Requests for Startups." Their framing: agencies of the future will look more like software companies, with software margins. And the numbers back it up. Agencies that seriously implement AI in delivery are seeing gross margins shift from 40-50% toward 70-80%. Delivery times compressing 3-5x.
JackandAI fits that description. We don't have legacy rate cards to cannibalize, no overhead structure to protect. That's not because we're smarter. It's because we started after the rules changed.
But Y Combinator's framing has the same blind spot as most of the industry. It treats the transformation as a margin story. Faster, cheaper, leaner. DEPT calls it "content as infrastructure." They're right that AI without a system behind it accelerates fragmentation, not scale. But even that framing stays inside the efficiency paradigm.
The safe middle (predictable creative at predictable cost) is exactly where AI commoditizes fastest. A feed full of content that performs and means nothing. Brands that speak constantly and say nothing.
Utendahl saw it. The margin story is not the story.
What I didn't anticipate
I thought the hard part would be the technology. It wasn't.
I thought the hard part would be team structure. Harder than expected, but manageable.
What I didn't anticipate: how difficult it is to keep the soul of what you're doing visible when you're building systems. Systems don't have souls. They have inputs and outputs. And the longer you spend building them, the more natural it becomes to think in inputs and outputs.
Azeem Azhar calls this the difference between cognitive offloading and cognitive surrender. Offloading is strategic: you let AI handle what it handles better. Surrender is when you stop reasoning altogether and let the system decide what matters. Most agencies think they're doing the first. They're drifting into the second.
I catch myself doing it too. Last month we were reviewing output from a content analysis tool, and I realized I was accepting the recommendations without interrogating them. Not because they were wrong. Because the system was fast and the results looked clean and I had twelve other things to do. That's how surrender works. Not as a decision. As a default.
The thing that actually matters
The answer I keep coming back to is cultural intelligence.
Not data. Not trend reports. Not AI-generated insights packaged as strategy. The thing that a good creative director does when they walk into a briefing and read what the room isn't saying. The thing that happens when you understand a brand well enough to know when not to follow the algorithm.
Let me be precise about this, because "soul over data" is a lazy argument.
The performance models we're building at JackandAI are genuinely useful. Pattern recognition finds things a human art director would never articulate. Videos with similar popularity trajectories predict each other's engagement better than content-similar videos. A cooking video and a dance video can follow the same viral curve. The model sees things we don't.
The problem is when the model says "this will perform three times better" and nobody in the room has the authority, the vocabulary, or the structural permission to say "but this isn't us."
At JackandAI, the cultural intelligence layer sits above the optimization layer. Not below it. The model is an input to human judgment, not a replacement for it.
That sounds obvious when you write it down. It is extraordinarily difficult to maintain when you're building the systems, running the sprints, and the optimization layer keeps producing results that look like progress.
Values claims are harder than org charts
Madison Utendahl walked away from a winning agency because she'd lost sight of why she was winning. The awards, the revenue, the recognition. None of it answered the question she'd stopped asking.
I'm trying not to stop asking it.
What JackandAI is actually building is a way of working where the cultural instinct (the thing no marketing curriculum teaches, the thing Zoe Scaman calls the competency that marketing's calcified canon cannot test for) sits at the center, not the edges. Where AI makes the execution faster and the analysis sharper, precisely so that human judgment has more to work with, not less.
That's not a structural claim. It's a values claim.
And the honest truth is: I don't know yet if values claims survive contact with the market. I know the margins work. I know the structure works. What I don't know is whether a five-person agency can scale cultural intelligence without diluting it. Whether the thing that makes it valuable is also the thing that makes it fragile.
The soul is not a luxury. It's the only thing left that's genuinely hard to copy. But it might also be the hardest thing to keep.