It's worth pricing that custom app again

I built an internal dashboard for our account team with Claude. It replaced a spreadsheet (surprise), and I was using the first version within a day.

That speed turned up something I didn't expect. For years, when one of our developers tore apart a set of wireframes over what looked like trivia, I'd wave it off. Radio button or checkbox, cards or a list, clickable or not: the client approved it, let's keep moving. Weeks later, in QA, I'd notice they were right, but now the timing, previous approvals, and impending launch made this more challenging to solve for. 

This time those same "trivial" decisions showed up while I was using the thing, not reviewing it, and because changing them was cheap, I just did. Claude didn't catch any of those problems, by the way. It built what I asked for, including the parts I (and it) had wrong. Knowing what to fix was still a human call, and usually one someone in Creative or UX would make.

This isn't only a lesson about wireframes. Clients used to ask whether we'd build their project agile, and I'd say no, we'll do waterfall. That cost us work, but it was the honest answer. Most clients have a problem, a budget, and a deadline, and paying an agency to iterate in sprints was expensive in terms of cost, calendar time, and their own time. So we scoped it all up front and worked hard to get it right the first time. Or at least tried to.

That's also why a lot of internal tools that clients brainstormed never got built. Next to a five- or six-figure build you had one real shot at, the spreadsheet was good enough.

We had our own version of this. A few years ago we mostly stopped proposing custom app builds, because they kept running over budget and schedule and annoyed everyone. We would rarely build them even for ourselves.

That math has changed. The first build is cheaper, and the even bigger difference is that the second and third passes are less costly too. A fixed scope and a fixed budget still makes sense. There's just room inside them now to be wrong and change course, which was always the expensive part. 

Cheap, fast, and good pick two: has long been a saying in our industry but cheap iteration doesn't mean you can skip the thinking. I could fix my dashboard by using it because I was the user and I knew what I wanted. Most tools are for people who aren't in the room, and they don't love change, especially when something that used to be easy gets harder. We use AI for a lot of the build now. But we didn’t rely on it to create our new website. We still do discovery and ideation, and we still make wireframes and designs. The difference is that we can start from a rougher idea and get something in front of real users sooner.

So if something important at your company still runs on a spreadsheet because the custom version didn't fit into the budget a few years ago, it's worth asking again. Both halves of the answer have changed: what it costs, and how right you have to be before you start. (Ask us, ideally. But ask someone.)