Nobody Asked for Sparkles

I built this site with AI. Not as a gimmick, but because Claude and Cursor let me do in two months what would’ve taken me a year to learn on my own. I’m not writing this as someone who’s skeptical of the technology. I use it daily, and it’s changed how I work for the better.

The upside is real, and not just for me personally. Product teams can explore more directions, faster, with fewer people, and ship functionality that helps customers. That’s not hype. I’ve seen it work first-hand, and believe that we should continue investing in it.

But there’s a side effect showing up across the industry that has nothing to do with helping customers, and everything to do with being seen to help them. This phenomenon is worth naming. Let’s call it AI signaling.

Here’s the problem. When AI makes a product better, it usually just looks like the product got better. Customers get more of what they want, faster. That’s the whole point. But “the product got better” doesn’t move a stock price the way “we shipped AI” does. So what you get instead is a wave of AI features nobody asked for, that customers never demanded, and that from what I’ve seen, provide little real use.

I’m not speaking from a distance. I watched this happen twice, a decade apart, from the inside, at two very different companies, in two very different eras of the AI conversation.

The first time, product teams were asked to find a way to incorporate a flagship AI platform into whatever we were building, not because it solved a real problem, but because leadership needed to point at usage numbers. We pushed back. The obvious question was always the same. If it doesn’t make the product better, why are we doing this? The answer, unspoken but understood, was that the market needed to see it everywhere. Not because customers were getting more value. Because it needed to be there.

Years later, at a different company entirely, when AI investment exploded industry-wide, it felt like déjà vu. The same internal conversations. The same designers and engineers agreeing, among themselves, that a feature didn’t make sense, while still being asked to ship it. Sparkles. Fly-outs. Extra context nobody requested.

I see the same pattern from the outside now too. Search on Amazon recently, and instead of a list of products, you’re dropped into a chat interface meant to help you find what you want. I can see the version of this that works, where AI makes the results list smarter, more personalized, more relevant, without you noticing the mechanism at all. What I actually get instead is an interruption. I didn’t come to have a conversation. I came to buy something. The chat isn’t adding value. It’s standing between me and the thing I wanted to do.

Here’s the point underneath all of this. Customers don’t care about your AI. They care about the product being better, faster, and easier to use. Nobody is grateful for sparkles and a chatbot that interrupts their flow. If anything, we should question whether features like this improve the experience at all, or degrade it while the metrics that matter to leadership go up.

I don’t say any of this from a position of naivety about why it’s happening. Companies have poured enormous capital into AI, and shareholders expect a return on that investment to show up somewhere visible. I understand the pressure. I just believe that the current response to that pressure is making our product experiences more bloated and less functional.

The better answer is the boring one. Build the thing that makes the product better. Let AI do that work in the background, and only bring it to the forefront when it’s an earned position. If it’s good, customers will feel it, even if they never notice the technology behind it. That’s a harder story to tell a shareholder in one slide. It’s also the only version of this that respects the customer on the other end of the product. Watch for AI signaling instead, features that exist to be seen rather than used, and you’ll start noticing it everywhere.