There's a question I get asked more and more lately. It comes after getting to know each other a bit better and is usually delivered with a slightly mischievous smile:

So… aren't you a lot faster now, with all this AI?

What the question really means, of course, is: could this be cheaper?

My answer usually earns me a raised eyebrow first — and then, almost every time, an accepting nod. Because my answer is: yes, we are faster. And no, that's not the point.

This post is my attempt to unpack that answer properly. It's about what AI actually changed in how we work at Dinghy, what it didn't change, and why I've come to believe that the interesting effect isn't that the work gets faster. It's that it gets better.

Let's name the fear first Link to this headline

Before I make any optimistic argument, I want to acknowledge something honestly: the anxiety around AI is real, and it's not irrational.

According to Mercer's Global Talent Trends research , 40% of workers worldwide now fear that AI will make their job obsolete — up from 28% just two years earlier. The headlines don't help. Layoff announcements get attributed to AI whether or not AI had anything to do with them — even OpenAI's Sam Altman has pointed out that "almost every company that does layoffs is blaming AI, whether or not it really is about AI." If you're a designer, a developer, or a copywriter, you've probably had at least one quiet moment of wondering whether your craft is being automated out from under you.

I won't pretend that fear away. But I've been watching what actually happens inside our studio and inside our clients' companies over the past two years, and what I saw leaves room for optimism after all.

PwC's global research on AI-exposed industries found something remarkable: the companies adopting AI most intensively saw productivity rise dramatically — and instead of using those gains to cut costs, the top adopters were growing headcount and wages. PwC's own conclusion was that AI may turn out to be a job expander, not a job killer.

The iceberg below every project Link to this headline

Every digital project is an iceberg.

The tip — the part above the waterline — is what clients see and get excited about: the brand, the interface, the animations, the launch. It's what gets presented in meetings and shared on LinkedIn.

Below the waterline sits everything else. The technical scaffolding nobody ever sees. Accessibility. SEO groundwork. Responsive behavior across dozens of screen sizes. Cropping the same image into fourteen formats. Working real copy into wireframes so a client can actually imagine the result. Login forms. Cookie handling. The long tail of small, necessary, invisible tasks that no client ever assigned value to — and rightfully so. They just expect this stuff to work.

Here's the uncomfortable truth every agency knows:

The iceberg's underside is where budgets go to die.

Every agency in our partner network tells the same story. The project starts with ambition — this time we'll do the custom shader animation, the delightful micro-interactions — and then the long tail eats the timeline, and by the end, the ambitious ideas are the first thing cut. "Let's see how far we get" is the sentence that quietly kills more good design than any client feedback round ever did.

This is precisely where AI changed our work. Not by replacing design or engineering judgment — but by eating the underside of the iceberg. The scaffolding, the format variants, the boilerplate, the long tail. The work that was always necessary and never meaningful.

The underside of the iceberg is where budgets used to die. AI-assisted workflows now absorb most of it.

We've watched this play out on real projects, not just in theory:

When we helped our friends at PAZE on a late marketing website project earlier this year — migrating hardcoded content into a proper CMS, the kind of unglamorous, error-prone grunt work that eats a week and produces nothing anyone will ever compliment — an AI-assisted coding workflow did it so cleanly it's now how we approach every migration like it. What used to be days of work is now not even twelve minutes. Though I should be precise: those twelve minutes presuppose that the tooling is running, that somebody knows what a content schema is, and that somebody knows what to type. The prompt is the visible tip. The know-how underneath is, once again, the iceberg.

Our own image tool started the same way: cropping one picture into fourteen output formats for a dozen channels used to be exactly the kind of task nobody budgeted for and everybody resented; now it's a drop-and-go step in our workflow.

My personal favorite productivity boost happened in our currently ongoing project, a franchise-network platform we're currently building for a client: In about eight weeks — mostly spent on discovery and concept work together with the client — we went from a blank page to a real, working web application: not wireframes, an actual product we could put in front of their partner network at an annual conference.

Before AI, that scope in that timeframe wasn't a stretch goal, it simply wasn't on the table. And the value didn't stop at the demo: we're now rebuilding that same software for production over roughly half a year, and because the prototype is real, working code rather than a picture of an idea, we're able to reuse a large share of it directly rather than starting from scratch.

It was never the tool Link to this headline

I want to be honest about why that worked, because it wasn't the tool.

It worked because we closed the gap between design and development years ago. We design in HTML anyway. We think in interfaces and UX patterns anyway. Daniel makes things work while I make them usable, in the same codebase, at the same time. Skipping the Figma step, the classic wireframe-design-implementation chain, and designing directly on the living product was a leap of faith. But it's a leap you can only take if that gap is already bridged.

Hand the same tools to a team where design and engineering still talk through handoff documents, and you get the same handoff documents, slightly faster.

That is the part the eyebrow-raise misses. The speed isn't in the tool, it's in the way of working the tool landed on. And once the long tail is handled, the economics of a project change in a way that has nothing to do with discounts. The honest version of my answer is this:

for the same budget, we can now spend dramatically more of our time on the parts that actually determine whether a project succeeds — strategy, concept, research, and craft. The things we always wished we had more time for. If anything, now is the moment to invest a little more, because every hour buys more than it used to.

What becomes a commodity — and what absolutely doesn't Link to this headline

I sometimes say, half-provocatively, that design and code production are becoming commodities. I know how that sounds to fellow designers and developers, so let me be precise about what I mean — because the nuance is the whole point.

There's a divide opening up between service providers right now. One camp keeps its prices where they were, makes the team smaller, and uses AI to deliver faster. Forrester's 2026 agency survey found that 81% of agencies now cite productivity as their primary reason for adopting AI, and warned that this efficiency focus is undermining creativity and long-term brand growth. That's the commodity trap: treat AI as a discount machine, race to the bottom, and hope your clients don't notice that everyone's output now looks the same. The other camp, which is where we've landed, points the automation at the mundane long tail, the stuff that actually eats the hours, so there's more time for the real brain work. The first strategy feels great for a quarter. You can run like that for two or three months. Then everyone is done.

But our experience points somewhere more interesting. Production isn't simply "commodity" — it's commodity, conditional on an upfront act of senior judgment.

Take code. If a senior developer sets up the architecture, the conventions and the guardrails right, AI agents can work productively inside that structure for a very long time. If that judgment is missing at the start, the codebase blows up, and it reaches a point where even the most capable models can't maintain it anymore. We have the comparison in-house: one codebase that grew without guardrails and is, to put it kindly, a burning dumpster fire, and one built deliberately on that senior judgment. Same models, same people, same year. The difference is the whole argument.

Design works the same way. Prompt an AI for a layout, a logo, a color palette, and you'll get something pretty good. Reliably, impressively, pretty good. But never excellent, and never genuinely new — because these systems, by their nature, give you the statistical middle of everything they've seen. Excellence and novelty still come from somewhere else: from a senior designer who has seen enough to judge what good actually is, and why — and who can define the system, the constraints, the taste that everything downstream inherits.

Notice what changed and what didn't. A designer who used to push anchor points in Figma now prompts. That doesn't mean they're no longer responsible for where the anchor points go. The task is the same; the means moved.

There's a word for this that the industry is rediscovering: taste. The AI in Design Report (Designer Fund / Foundation Capital) put it well — the designers proving most valuable right now aren't the fastest adopters, but the ones who kept the clearest point of view on quality while everyone else optimized for speed. Taste is what senior experience is. It's the same thing as the senior developer's architectural judgment, just in a different material.

So here's our working model, in Pareto terms: seniors invest their 20% of effort where it produces 80% of the outcome — setting the strategy, the system, the guardrails, the taste. AI then handles the long tail underneath. Not the other way around. The agencies getting this backwards — using AI to generate the concept and humans to clean up the production — are the ones Forrester is warning about.

Senior judgment first, AI inside that structure. Getting this backwards is the commodity trap.

If you're a designer or developer feeling the fear: your craft isn't the commodity. Your taste — the accumulated judgment of everything you've seen and built — is becoming the scarcest resource in the room. Lean into it.

So yes, we are faster. But the honest answer to the eyebrow-raise was never about speed, and the biggest opportunity in all of this isn't inside our studio at all. It belongs to our clients, and to the people in their teams who never get to stop. That's the second iceberg, and it gets its own post.