"Should we be adding AI to what we do, too?" Over the past few months, startup founders have kept circling back to this question, and the backdrop is always the same. Look anywhere in the company deck or the investor pitch and you'll find a line about being "AI-powered," offering an "AI solution," or having "adopted AI"—yet far more often the term is dropped in like a buzzword than built into the product as a core feature. In a recent piece for Outstanding, business consultant Lee Bok-yeon argues that the question itself is the problem. Before you go looking for an answer, he says, you have to change the question.

At first it sounds a little severe. Calling it "the wrong question" can feel like it brushes off a very real, on-the-ground worry. But follow the argument and it becomes clear: this isn't a case against using AI. It's a case for going back and asking where the thinking should actually start.

The more decks say 'AI,' the more they start to look alike

For the past few years, finding the word "AI" in a startup deck or pitch has been the easy part. The harder thing is finding a deck without it. SaaS tool, commerce platform, content service—each one works in at least a line about AI. "AI-based recommendation algorithm," "AI automation," "AI-enhanced customer experience": the wording varies, but the shape is the same.

Several forces are at work behind this. Investors ask about AI. Competitors lead with AI. No job posting leaves out AI experience. The reasons differ, but they converge on the same outcome—the anxiety that "leaving AI out will make us look like we've fallen behind" works its way into the deck.

The trouble is that, in the process, the decks start to resemble one another. What problem the company is actually solving and for whom, why this team is the one to do it—that detail fades, and "using AI" moves in to fill the space. An investor is trying to read what sets one business apart, but when everyone reaches for the same word, the differences get harder to find. You add the word AI and, paradoxically, the company's own color gets erased.

Lee's essay puts its finger on exactly this point. "Should we do AI?" is a question about choosing a tool. But to choose a tool, you first need a problem to solve. When the tool comes before the problem, the direction of the business ends up chasing whatever the tool of the moment happens to be.

The counterargument: 'just slap it on and fill it in later'

There are people who flatly disagree, and their strongest objection is pragmatic. "Claim AI first, then fill it in later."

The numbers lend this view some support. From 2023 to 2025, AI-related startups took up a fast-growing share of total venture investment, and it's become an open secret among practitioners that without an AI label, you can't even land the first meeting. If the market wants AI, then attaching the label first as a survival strategy isn't entirely irrational.

There's another objection, too: "the label pulls the direction along." There's no shortage of teams that declared themselves AI and then actually went and built the capability. The experience of announcing first and then moving toward it, the argument goes, can even produce faster execution.

But for this to work, one thing has to be in place. After claiming AI, there has to be a "why" that actually moves the team in that direction—a reason for which problem you intend to solve with AI, and why. Without that reason, the team starts to lose steam in the gap between what it says and what it does. An AI label may open the door in the first meeting, but in the second meeting the investor asks about that gap. Having to repeat "we're still building it" can cost far more than the early advantage was ever worth.

For a solo operator, this question arrives in a different form

This may sound like a story about startup pitch strategy, but for a solo operator or a small team the same structure shows up differently. Not as "should I use AI," but as a vague unease: "if I'm not using AI tools properly, aren't I falling behind?" That unease leads to guides, courses, newsletter subscriptions—and at some point you're spending more time scouting AI tools than doing the customer work you actually need to do.

The more limited your resources, the more this cost stings. Sometimes the time it takes to learn a new tool well enough to actually use it is greater than the time the tool would ever save you.

In this situation, I'd argue that changing the order of the question is by far the faster route. Not "should I use AI," but first: "what is the task I repeatedly spend the most time on right now?" Once you have a concrete answer to that, the next question narrows on its own. Is there a way to cut down that repetitive task? If so, does an AI tool work for it in practice? And is the time it takes to learn smaller than the gain?

Order it this way and choosing an AI tool gets much easier. Instead of the pressure to try everything, you can focus your scouting on the single task that's eating the most time right now. Narrowing the search to one thing—deciding, say, that this month you'll only use it for drafting. The narrower the scope, the lower the learning cost and the easier it is to see whether it works.

There's one more distinction worth drawing. Does the AI tool make something you're already good at faster, or does it handle something you're not good at on your behalf? In the first case, your existing strength gets sharper. In the second, it's hard to build real capability in that area, and when the tool changes or disappears you're left wobbling along with it.

Some people follow the trend flawlessly and still come away less sure of what their business even exists for. Chasing the format of the moment, the keyword of the moment, the tool of the moment, until one day they can no longer explain what color their own brand is. And looking back, the businesses that last tend to be the ones that were dogged about the problem they were trying to solve, not the ones that were most sensitive to trends.

"Should I use AI" isn't a matter of right or wrong. It's just that the harder you try to answer it first, the less time you spend asking where you're actually starting from—and Lee's point is a proposal to flip that order. Once the problem comes into focus, the tool is something you can choose next.