Just AI It

Everyone wants an AI Product. Few stop to ask what it means to design one – and why the answer is older than the technology.

Design used to be simpler to explain.

In the 90s, designers made things look good — posters, brochures, packaging, and eventually the first websites. The job was pixels. Then the web grew up, and “designer” started meaning someone who made things work — banners became interfaces, and interfaces became products. By the 2010s, we were UX/UI designers: user flows, wireframes, usability tests, design systems. We weren’t decorating screens anymore — we were orchestrating how people moved through them. And somewhere in the last few years, the job stretched again, past the screen and into product thinking and strategy.

Now there’s another shift. Designers aren’t just pushing pixels or drawing flows, we’re prompting, curating, directing. AI can generate ten variations of a screen before your coffee gets cold, synthesize a hundred user interviews in an afternoon, draft a working prototype from a sentence. The task is evolving from making the thing to deciding what’s worth making — and guiding the machine toward it.

Which might be why every request now sounds like: “Just AI it!”

Can’t decide on a feature? AI it. Users not converting? AI it. Competitor launched an AI chat box? We need one too.

And look – I get it. At Base87 Technologies, we build AI-powered tools ourselves. One example is Dayline Observer, which digests daily news updates in Hong Kong, AI made it possible in a way they weren’t before. I’m not an AI skeptic. I’m not an AI evangelist either.

I’m somewhere in the blur again – and after a couple of years of building, explaining, and occasionally untangling AI features, I’ve realized something:

Focusing on user needs is still more important than chasing a new technology. The tools changed, but the job didn’t.

This article is my attempt to answer the question clients and teams keep asking me – how do you actually use AI in your work? – in two parts. Part one is about the craft: the pitfalls that weaken product decisions, and the virtues that strengthen them. Part two is about strategy, and why AI doesn’t replace product thinking.

The goal was never to be AI-first or AI-avoidant. It’s to be value-first, problem-first – and let AI earn its place in the solution.

Part One: The Craft — Avoiding Common Pitfalls in Product Work

This part is for product teams: designers, PMs, engineers, anyone who makes stuff that people use. Because how we work with AI day-to-day either sharpens our decisions or quietly erodes them.

The AI Pitfalls

Shipping the demo, not the product. AI demos beautifully. It also fails confidently, inconsistently, and at the worst possible moments. If your feature works 80% of the time, that’s not a product – that’s a party trick. The pitfall is designing only for the happy path and treating error modes as someone else’s problem.

AI-washing. If a dropdown menu solves the problem, build the dropdown menu. Slapping “AI-powered” on every chance or feature possible undermines trust and misleads users, teams and stakeholders.

Letting the model make the judgment call. AI is excellent at generating possibilities. It is not excellent at knowing which one is appropriate for your users, your business, your context. The moment a team outsources its judgment to the model, it stops being a product team and becomes a prompt team.

Skipping research because “the AI will learn.” No. A model learning from usage data is not a substitute for you understanding your users. Discovery, interviews, watching people struggle – none of that gets automated away. Teams that skip it just fail faster, with better tooling.

Designing for the technology instead of the person. The most common pitfall in a new form. We did it with mobile, remember ‘we need an app!’? The technology changes; the pitfall somehow is the same.

The AI Principles

Start with the problem, not the model. Before asking “what can AI do here?”, ask “what is actually painful for our users?” Sometimes AI is the answer. Sometimes it’s a distraction. You only find out by starting with the pain points.

Use AI to understand users better – not to avoid understanding them. This is where AI genuinely shines for UX practitioners: synthesizing research, clustering feedback, spotting patterns across hundreds of support tickets in an afternoon. It doesn’t replace your judgment. It gives your judgment better raw material.

Prototype faster, validate sooner. AI collapses the cost of exploring ideas. Use that. Test five directions in the time you used to test one – then let real users tell you which one deserves to exist.

Design for trust. Show your work. Let users see why the AI did something. Give them an undo option, and keep a human reachable. Confidence indicators and clear ‘I’m not sure’ moments are crucial parts of the UX in AI products.

Measure outcomes, not AI usage. “We added AI” is not a metric. Did task time drop? Did errors fall? Did retention move? If not, the AI is decoration.

Part Two: Why AI Doesn’t Replace Product Thinking

Even AI-powered products require a deep understanding of your users – their use cases, their unmet needs, their mental models. Maybe especially AI-powered products.

Because here’s the truth nobody says out loud:

People don’t adopt an AI feature because of the feature itself.

They don’t care about interacting with a large language model for its own sake. They use your AI feature because it removes a pain, saves them an hour, or gets them to done with less effort. When we built Dayline Observer, a Base87 Technologies project, the goal wasn’t to showcase AI — it was to tackling the real problem of information overload. Teams were drowning in updates, and we wanted to build something that could read everything so they didn’t have to. The AI is the plumbing. The value is the water.

Here are a few uncomfortable implications:

Your users’ mental models still run the show. People bring expectations shaped by every tool they’ve ever used, including every frustrating chatbot that ever misunderstood them. If your AI behaves in ways that violate how people think the task should work, no amount of model quality will save you.

“It’s AI” is not a moat. Your competitors have access to the same models you do. The differentiation is everything wrapped around the model: your understanding of the workflow, the trust you’ve built, the UX that makes the product output actually usable. That wrapping is unglamorous product work.

The unmet need comes first; the technology earns its way in. The strongest AI features I’ve seen started as a clearly articulated user problem – and AI happened to be the first technology that could finally solve it well. The weakest ones started as “we need an AI story” and went looking for a problem to justify it. Users can tell the difference, even when they can’t articulate it.

So when the pressure comes – from leadership, from investors, from a competitor’s announcement – the answer isn’t “AI it” and it isn’t “never AI.” It’s the same questions it has always been. What job is the user hiring us to do? Where’s the unmet need, and how painful is it really? Does AI genuinely change what’s possible here – or just what’s marketable?

If AI answers those questions well, use it fully and proudly. If it doesn’t, no roadmap will make it so.

The Request Changes. The Responsibility Doesn’t.

The tools keep changing, and the requests keep changing with them. But the job of a product team has not changed: understand people, solve real problems, and take responsibility for the outcome.

AI is a remarkable addition to the toolbox. It can make us faster, sharper, and more ambitious about what’s worth building.

What it can’t do – what it will never do – is care about the user on our behalf. That’s still our job.

If you’re navigating your own version of “Just AI it” or building something where AI genuinely earns its place, I’d love to hear what you’re working on. The best conversations always start with someone willing to ask: “What’s the actual problem here?”

Thanks for reading! 💗