Yes, And We'll Figure It Out

We sat down as a team recently and recorded a long, unscripted conversation about how AI has changed the way we work. Not the keynote version, with the clean arc and the tidy lessons. The real version, with the wins, the open questions, and the parts that still make us a little nervous. What follows is some of what came out of it, lightly organized so it holds together on the page.

What we can say yes to now

We've always had it in our DNA to say yes and figure it out. What changed is the cost of "figure it out."

Two projects make the point. A few years ago, a large interactive install for a major manufacturer took a team of eight to ten people and most of six months. Designers, front-end developers, backend engineers, six figures of budget, and a result that looked great up front while being held together with tape behind the scenes. Recently we built something in the same ballpark with two people.

The tools are part of that. The bigger change is permission. We can now test an idea in 24 hours to learn whether it's even true, then commit with our eyes open. The difference is that we're not just promising something and sorting it out later. We can actually go validate the thing in a day.

You can watch it change individual people, too. Eighteen months ago, Dylan was mostly our video guy. The idea that he would cheerfully take on a custom interactive build with a hard six-week deadline would have sounded reckless. Now it's a normal week. On one recent project we prototyped a game concept in an afternoon, right after the brief, while the traditional process on the client side wouldn't have a working version for two months. We didn't even end up needing to show it. Having it changed how we thought about the whole job.

The catch nobody warns you about

The same tools that let us say yes are also flattening everyone's work into the same look.

A client recently ran some of their writing through an AI tool and told us, themselves, that it felt "very Claude-y." Clean, competent, universally likeable, and completely generic. A different client pointed an AI tool at the brand guidelines we had built for them and got back a polished deck that was quietly off-brand: invented brand colors, serif fonts they had explicitly rejected, a soft cream where their crisp white should be. It looked nice. That's the trap. If you don't have a designer's eye, "nice" is exactly the level that fools you into stopping. It clears every bar you know how to check, while quietly failing the ones you can't see: distinctiveness, consistency, and whether this could only have come from you.

Brand is consistency over time, and generation is the opposite of memory.

Consistency is the part these tools are worst at. You can make one really good post. Go back the next week for the second in the series and it comes out close, but not the same. The spacing drifts, a color is off, the logo sits a little differently. I came up more on the strategy side than the design side, and even I can spot the drift instantly, because I've been staring at a given brand for months. Holding that standard and applying it every single time is unglamorous work. It's also hard to automate and easy to underestimate.

So why hire anyone, then?

Which brings up the question we ask about ourselves. If anyone can type a sentence and get an image, a video, or a working app, why hire an agency at all?

Part of the answer is time. Most of our clients could learn our tools. What they don't have is the hours to live in them daily and watch them change weekly. The other part is taste, and that's not going anywhere. Underneath every deliverable, an agency does one thing: we validate. We say this is on brand, this will land, this is worth your name. The tools changed how fast we reach a draft. They didn't change who decides the draft is right.

Nacho on our team framed it well. When a client brings us something AI made and likes it, part of our value is being able to say why they like it, and what's actually working underneath. Taste, made explicit and useful.

Where this is heading

A few threads we are watching.

One is a question I keep poking at: who is your website even for, now that fewer people visit it? More and more, your customer isn't on your site. They're asking an AI about you, and the AI reads the page so the human doesn't have to. So what happens if you design for the agent, a ruthlessly clean, well-structured page built to perform when a machine is the reader? At that point, as someone on the team put it, the homepage is basically a tidy text file. Beautiful design isn't going away, and in-person and visual experiences are arguably getting more valuable, not less. But there may be a whole machine-facing layer of the web that the smartest brands start optimizing on purpose.

The other thread is a hunch about taste swinging back. The more everything looks the same, the more people will crave the weird, characterful stuff again. Think of the early web, GeoCities and Angelfire, ugly but designed by a person with a point of view, the kind of page you actually remembered. You can already see the equivalent in film, where cheap, flawless cameras pushed everyone toward an identical clean look, and now people add grain and imperfection on purpose to break out of it. Once flawless is free, flawed-on-purpose becomes the flex.

A multiplier, not a replacement

That was the phrase we kept circling. These tools don't replace the designer, the writer, or the strategist. They multiply what each of them can attempt in a day. They also reward being in it daily. Aaron said he can't picture his workflow without them now, and I agree it would be tough to step in cold today and grasp what's possible. It's exciting and a little terrifying at the same time, and it's a good time to be scrappy.

One last bit of perspective, to keep us honest. When we looked into how many people actually use these newer tools day to day, the answer is: not that many yet, next to the millions who have tried a chatbot once. Enterprise friction is real, and most people simply aren't moving this fast. So either we're a little ahead of the curve, or we're drinking the Kool-Aid. Probably some of both. But every brand we work with will eventually have to make these tools part of how they operate, and when they do, they'll want a partner who has already taken the arrows and can hand them the version that actually works.

That's the job now. Same as it ever was, really. Good taste, applied with care, on behalf of people who have better things to do than keep up. The tools just made the "yes" a lot easier to say.