AI Is Not a Feature, Stop Treating It Like One

“Can you add AI to our app?”

Sure. But what problem are we solving?

“We just want AI in there.”

We’re having this conversation more and more, and it’s not just startups. It’s established businesses, big brands, companies with real engineering teams. They want AI. They just can’t tell you why.

AI is a category, not a feature

The thing people seem to have missed: AI isn’t a feature. It’s not something you add like a dark mode toggle or a search bar. “AI” is a category, as broad and varied as “software” itself.

Are you talking about a recommendation engine? A chatbot? Automated document processing? Predictive analytics? Real-time personalisation? Generative content? Anomaly detection?

These are completely different technologies, with different costs, different risks, and different levels of effort. Lumping them under “AI” and asking someone to build it is like walking into a hospital and saying, “Can you do medicine on me?”

Box-ticking dressed up as strategy

What frustrates us most is the box-ticking. Companies that want to tell the market they have AI, before they’ve identified a single problem it would actually solve. That’s not innovation. That’s branding dressed up as strategy.

The conversations tend to follow a familiar pattern. Someone in the business reads an article, attends a conference, or hears a competitor mention their “AI-powered” product. Within days, there’s a brief on the table. The brief has no defined problem, no success criteria, and no real budget, but it has a deadline because the quarterly report goes out in six weeks.

This is how you end up with an AI chatbot that no one uses, a “smart” recommendation engine that recommends the same three products to everyone, or a generative content tool that produces text your team has to rewrite anyway.

Start with a boring question

The best AI implementations we’ve seen start with a deceptively simple question…

Where are people in your business doing repetitive, time-consuming work that a machine could handle better? Or: Where are users dropping off because the experience is too slow, too generic, or too manual?

Find that. Then figure out if AI is even the right tool. Sometimes it is. Sometimes a better process, a faster database query, or a well-structured form will solve the problem faster, cheaper, and with far less risk.

The question that saves months

AI is only as useful as the clarity behind the decision to use it. If you can’t explain the problem, you can’t evaluate the solution. And if you can’t evaluate the solution, you’re going to spend a lot of money finding out it doesn’t work.

So next time someone in your business says “we need AI”, push back. Ask them what, specifically, is broken. That single question will save you months of misdirected effort and budget you won’t get back.

If you’d like help asking the right questions before you build anything, LET’S CHAT!