Three Chat Bubbles

Recently, I looked at our product and saw three circles in the bottom-right corner of the screen. Each one let you send a message.

There was a reason for each. We work in post-acute care, where admissions teams need to talk internally about a referral, communicate with hospital case managers, and coordinate across facilities.

Three useful capabilities. But looking at them together, I had a very basic question: which one am I supposed to click?

That’s something I’ve been thinking about a lot lately.

I used to work through a pretty linear design process. Look at the data, talk to users, sketch, wireframe, prototype, get feedback. Eventually, hand it over to engineering with enough detail to avoid expensive mistakes.

Development time was precious. A lot of our process existed to protect it.

Then AI flipped the equation.

With the context of your codebase, you can go from an idea to something that actually works, with your actual data, incredibly fast. That’s amazing. It changes when and how you can learn.

You can put something real in front of people and figure out whether it makes sense. You don’t always need several rounds of increasingly realistic representations before you get there.

But it looks finished before the thinking is finished. It has the right buttons. It connects to the database. It works. It’s easy to treat that as a feature that belongs in the product.

The three chat bubbles made this visible to me. Building each capability had become easier. Making the whole experience understandable still needed care.

Customers have to absorb everything we add. They have to discover it, understand it, remember where it lives, and figure out how it fits into their work. We can build faster than they can do that.

At some point, “we need to educate customers” deserves another question: how much are we asking them to learn?

This is why I’m spending as much time as I can talking to customers. Getting the vibe check.

Are they engaged? Are they giving specific feedback and suggesting ways to make it better? That tells me something. Silence is harder to read.

Then comes the judgment: which requests point to something that would help across our customers? How does that need fit into what already exists?

That work feels more important now.

So does being willing to throw things away. Our team is up for that, and I think it’s critical. If we can build something quickly, we should be willing to remove it quickly when it isn’t helping and is making the experience more cluttered.

Sometimes that means deleting a feature. Sometimes it means keeping three useful capabilities and finding one clearer way to reach them.

AI gives us room to test sooner, learn faster, and spend more time making things feel simple and thoughtful. Some early versions should reach a few customers while we’re still figuring them out. Some should disappear.

We don’t have to spend every bit of time we save on building more.