There isn’t much that gets ideas flowing than seeing something you’ve designed up and running. Not a mockup of it, not a slide about it but the actual thing itself, out in the wild doing what it was designed to do, in use by the people it was designed for.
As the designer of such a system, application or website, you see things immediately. The action that could have been a single click but instead is three. Two data points sitting separately that when combined presents an entirely different story. These things are difficult to spot when you’re at the coalface, mired in the detail and become much more apparent when you see the thing in operation.
For the vast majority of the time we’ve been doing this type of work, those moments of realisation came far too late. You’d research, sketch, wireframe, design, refine, sign-off, build and only then, finally watch a real person use it, only to discover a raft of things you’d have changed if only you’d have known. Prototyping shortened that gap and conducting user research mitigates some of those later findings but both of these come at cost; both in time and actual financial cost. Prototyped user flows were rationed to ‘key journeys’ and user research might be reduced to a lower fidelity. The ideas you’re more sure about are the ones that get tested and the rest tend to be parked or added to the ever-growing backlog.
What’s changed?
We’re now testing thinking and ideas faster than we ever have before, by building the thing. A vague idea from a workshop can be a working screen the same afternoon, styled up and working with pseudo data. A ‘what if the dashboard led with this instead?’ can be multiple versions of the dashboard shown side-by-side instead of a debate based on vague abstractions of the same idea that might be held somewhat differently in people’s heads. We’re calling it hyper prototyping, a natural progression of the idea of rapid prototyping, which in our interpretation means something lesser, both in speed and scale.
What hyper prototyping means in practice is that the meandering, half-formed ideas also get tested and viewed at the same resolution as the ‘better’ ideas. The ideas that might never have left the whiteboard, considered unworthy of a days worth of production. Many of them, still inferior ideas, but we know it now instead of it being an assumption. A surprising number of the ones that survive however are better than the safe option that might’ve shipped.
This is where we’re seeing real gains from AI in our creative work. The tools aren’t having the ideas for us, they’re letting us look at more of ours and our clients own ideas and in much greater depth and resolution than we’d ever be able to previously and before anyone has committed to anything.
Faster, better and more
It’s tempting to see this as a speed gain, and it is. But what’s more interesting is what speed is doing to volume and what effect it is having on quality.
Invariably, every project has a fixed amount of time, what varies is how many ideas get a fair consideration. When a working prototype costs an afternoon instead of a week, the field of things we get to try gets much larger, and the final product is the survivor of that larger field. Doing it quicker doesn’t mean completing it sooner, it means more of the thinking gets finished and decisions are made on evidence not gut-feel. It also means inferior ideas or concepts are killed off earlier instead of limping into production because alternatives weren’t tested.
Getting rid of ideas matters as much as generating them. When a prototype takes a long time, people get attached to it and mediocre ideas survive on sunk cost. When it costs a few hours, it becomes easier to throw away.
What it changes for the people we work with
Reviews are different. Conversations about a design mockup end up being about how the thing looks whereas a conversation about a working thing is about how it behaves. Feedback and discussion around a working prototype can be had while experimenting (or being wrong) is still cheap. The next learning that we’re processing is how to manage expectations around a working prototype that feels finished. With the rise in sentiment that ‘I can just get Claude to do it’ it’s easy to fall prey to how impressive these tools can make things look whilst they remain a far cry from a secure, robust product.
This is the version of AI adoption that excites us, not a way of doing less and removing agency but a way of getting more of the thinking in front of people sooner so more of it is tested and less of it is assumed.
