For twenty years, the gating question on a software product was can we build it? Now, for almost any reasonable scope, the answer is yes. AI has flattened the build cost. What it hasn’t flattened is the part that actually decides whether the thing is good.

That part — the brief, the flow, the pixels, the way it feels in the hand — is what we’ve spent two decades doing.

01 — On scarcityThe bottleneck moved.

The pitch in the air right now is “AI agency.” That’s not us. We’re a design-led studio that uses AI the same way we use Figma, Photoshop, and a sharp engineering lead — as the tool that gets the idea out of our heads and into something usable.

What changed isn’t our craft. What changed is the throughput. A flow that took a sprint to prototype takes an afternoon. An integration that used to need an estimate now needs a decision.

Which means the scarce thing is no longer the build. The scarce thing is knowing what to build, where the friction should sit, what to leave out, and what the user should feel five seconds after they land. That’s UX. That’s design. That’s the part AI doesn’t do for you.

A team that’s spent twenty years staring at hierarchy, contrast, weight, and the small forensic question of why does this feel off? — that team gets a force multiplier.

— Field note · April 2026

02 — In practiceWhat we actually do with AI.

We build internal tools we’d never have justified the budget for. We ship client products faster, with the same polish we’d bring to a flagship site. We prototype against real data on day one, not week six. We test three versions of a flow when the brief used to allow one.

A few from the last stretch of work:

• i.
  Real estate · Dashboard
  A live sales bridge for one of our real estate clients that pulls Spark.re data into a dashboard their executive team actually opens. Not a vendor product. Not a templated portal. Built to their model, in their voice.
• ii.
  WordPress · Accessibility
  A custom WordPress plugin that auto-generates accessible, on-brand alt text for an image-heavy portfolio site — because the off-the-shelf options were either lazy or expensive, and the archive was too important for either.
• iii.
  Brand · Workshop
  A Brand Voice Workshop app that takes the exercise we’ve run in conference rooms for years and makes it work for any team, anywhere. More on that one in its own post.

None of those exist because we suddenly learned to write code we couldn’t write before. They exist because we already knew what should be there, and AI removed the reason we used to say no.

03 — On timingWhy this is the right moment for design-led teams.

Here’s the trade-off worth naming out loud: AI will give anyone the ability to make software. It won’t give anyone the ability to make software that feels considered. Those are different problems.

A team that’s spent twenty years staring at hierarchy, contrast, weight, spacing, and the small forensic question of why does this feel off? — that team gets a force multiplier. A team that’s never asked those questions gets a faster way to ship the same generic output.

Taste compounds. It always has. AI just made the gap visible.

04 — ForwardWhat this means for the work we make next.

Every product we build still starts the same way: with a question, not a comp. The first deliverable is still clarity — what’s the thing, who’s it for, what does it need to feel like, where do we let it be plain and where do we make it sing.

The difference is what comes next. We can answer that question in a working prototype now, instead of a deck. We can put the real product in front of the real client in week one. We can iterate on the feel, not just the spec.

That’s not a bigger toolbox. That’s a better conversation.