As we like to say at Ducky — when talking about your product, you don’t brag about the fact that you code with a specific language. It’s just a technology. Just a part of the stack a company uses to build software that customers want. AI should be treated in the same capacity. There’s no medal (apart from maybe a frothy fundraise…) for jamming AI into place left and right if the ultimate output doesn’t actually delight customers and help them accomplish a goal. Often times, folks start by asking “what can I build with AI?” instead of “what do my customers want to accomplish that can be uniquely improved with AI?”

We’ve had countless conversations with folks hell bent on building AI into their products. Many of these have been developers, but an equal percentage have been people who don’t code — product managers, sales leaders, operators, and executives. A misguided assumption we had going into these conversations, and this new product direction, was:

People know what they want to build. They just don’t know how to build it.

Turns out, a lot of companies don’t know what they want to build. They have the conviction that they must build with AI, or be left behind. Many have even received a strong directive from executives with the following guidance (and I quote): “we have to do something with AI.” But, for many companies, the question remains — what should be we build with AI?

At face value, this doesn’t sound like a wild question for a business to ask itself, nor one that should stump a team. It’s a slightly different permutation of the fundamental questions that most of us ask ourselves every day: What pains do our customers experience and how can we fix them?

And yet, we continue to notice an interesting, and counterproductive pattern — many teams don’t start with this age-old, user-centric, question. Instead, they ask:

• “What LLM works best for my industry?”

• “Is Claude better than ChatGPT?”

• “What chatbot provider performs the best for my sector?”

• “What agent-builder has the best pricing?”

At their core, the above are essentially vendor-selection questions. Questions that, when choosing to adopt new software for our companies, most of us ask after we’ve decided on what we’re trying to accomplish. You don’t take a demo with Brex before you need an intuitive corporate card solution. You don’t explore Superhuman without first thinking to yourself, “damn… this email backlog is killing me… there’s got to be a quicker way to do this.”

So, why do seasoned, successful folks in the world of business fall victim to this seemingly backwards way of thinking about AI implementations? A lot of it has to do with the sheer volume of marketing and buzz around the world of AI + its accessibility. Everywhere we turn, it’s “AI this” and “AI that” — it’s pollen in early April down here in Nashville — omnipresent and persistent. And, it’s not just a concept we see on our feeds but haven’t experienced first hand. Most of these products have generous trials or free tiers where we can actually get in and play around — and that’s wonderful. But, it also creates a dangerous thought pattern, specifically for those not deep in the code: “If I, as a non-technical person, can accomplish all this with AI… just think about the magic the dev team can create.”

The term magic is problematic in and of itself. None of this is magic. It’s software. Sure, it’s a tectonic shift in our previous perception of software, but still software nonetheless. AI cannot do everything. But it can do a lot of things pretty damn well. But, the difference between 60% efficacy and 90% efficacy is in the construction of the system you choose to create (or the tool you choose to buy to accomplish a specific goal).

AI supports us in the creation of truly incredible things, this is undoubtedly true, but don’t let the presence of AI distract from or cheapen the still-necessary exercise of critical thinking. The big questions are still the same:

• What pain am I solving for?

• What do we need to build in order to solve it?

Once you have a handle on the same ol’ questions, then it’s time to consider infrastructure, tooling and whether or not bringing AI into the mix can make previously impossible or operationally burdensome ideas a sudden, plausible, reality.

TLDR y’all — you’re better equipped for the AI age than you think. And, if you’re a bit lost, you’re in great company.