Now That You Can Build Anything, Build the Right Thing

Lately, every executive I meet asks one of two questions about AI. How do we add AI to our product? How do we use AI to ship features faster? Both miss the point. At a time when we could build anything, neither of these questions will guide you toward what to build. Neither protects you from spending a year and a million dollars on something nobody needs. The harder, older question still matters most: what is worth building in the first place? That question can't be answered by a model, a vendor demo, or a code-generation tool. It can only be answered by people who have learned to think clearly about value. And the smartest move a leader can make right now is to find a partner who has.

In my work with Atomic, I engage with potential clients early in their decision about whether to build their own software. I help business leaders think comprehensively about whether to invest in proprietary digital products. More often than you'd think, I recommend they don't. The risk is too great and the return on investment too murky.

So why does anyone build custom digital products? Because some leaders understand how to build for strategic market advantage. Digital products can change the destiny of entire firms or business units. They can give a company an advantage competitors cannot easily overcome. I'm not using hyperbole. Done right, software can truly change your entire business.

Take Microsoft Word. Development began in 1981. Word was released two years later (it was terrible, no one used it) but didn't gain broad use until 1993—twelve years of investment before any real return. How frustrated must Bill Gates and Paul Allen have been in 1987? Six years in, defending the product to the board, still spending, with six years still to go. They saw the value a premier word processor would bring. Today, 98% of the Fortune 500 hold a contract with Microsoft. The foot in the door, in nearly every case, is Word. Microsoft might not exist today without it. Instead, the company carries a market cap of $2.15 trillion.

Microsoft didn't win that fight by shipping faster. It won by building the right thing. Speed without strategy is just a faster way to spend money on the wrong product.

So how do you find the kind of value that lets you dominate your market? The answer has to do with weight and variance—and with finding someone who can help you see them clearly.

Understanding Weight and Variance Starts in School

(Hat tip to Prof. Sonia Marciano, who formalized these concepts for me.)

Weight

Back in 2021, I had a 15-year-old at home, and she struggled at school that year—she had never struggled like that before. Some of it was remote learning, but more of it was failing to grasp weight. She didn't see which assignments counted heavily toward her grade and which she could drop without consequence. She would spend all night on an assignment worth less than 1% of her final grade, then walk into a test worth 25%, drained and unprepared. She poured energy into things that didn't move the needle, while ignoring what mattered.

Businesses operate the same way. Some parts of your offering are crucial; others are practically irrelevant. Do you know which parts of your business carry weight? Which ones are crucial to the firm's success? How do your clients and competitors think about those same parts?

These questions are hard to answer on your own. The leaders who answer them well usually have help: a partner who has seen a hundred products in a dozen markets, who will push back when you say "let's add an AI assistant," and who will ask whether the assistant counts for 25% of the final grade or 1%. If it doesn't count, don't build it. Software should never be built for convenience; it should be built because your company cannot survive without it. If it doesn't count for at least 25% of the final grade, find another way.

Variance

I myself failed to understand variance in school. At my high school, final exams were graded on a curve, and I didn't grasp the value of the curve. A friend—who turned out to be our class salutatorian—once saw me sweating a chemistry final and asked why I was worried. "No one is going to do well on this final," he said. "Stay focused on the final project." That's when I realized: we would all get B's on the final and therefore A's. The real variance lived in the project, which wasn't graded on a curve. I stopped studying for the final, turned to the project, outscored my classmates, and ended with one of the highest grades in the class. Without that focus, I would have spent my time in the wrong place. There's a reason he ended up salutatorian, and I didn't.

Apply this to the business world. From 1981 to 1991, Word was a distant third behind Lotus 1-2-3 and WordPerfect. With the release of Windows 3, MS Office team leader Jeff Raikes identified the graphical user interface as the key area of variance. His team spent two years adding a point-and-click interface to Word while Lotus and WordPerfect chased niche features their customers didn't use. Raikes found the difference and executed. By 1991, Office had 90% of the productivity market and an 85% operating margin. Lotus was acquired by IBM in 1995 for a paltry sum and has since withered. WordPerfect was acquired in 1994. MS Office still prints money for Microsoft.

Variance is what Raikes had and what his competitors didn't: the ability to see, in a noisy market, which difference would matter. AI cannot give you that ability. A partner who has watched dozens of markets reorder themselves can. The wrong partner will help you ship the same chatbot every competitor is shipping. The right one will tell you that the chatbot isn't your variance—your data is, your workflow is, your domain expertise is—and help you build the thing only you can build.

The Choice That Matters

The best software partner will help mitigate some of the risks of building custom software. At Atomic, we take the stupid risk off the table. Your project will not fail because of the technology we build together. AI shortens some lines of code, but it does nothing to tell you which lines are worth writing. That work is human, strategic, and of the highest value.

Even the best partner cannot tell you alone whether your product carries strategic market value. That judgment belongs to you. But you don't have to make it alone, and you shouldn't. The most important decision in front of you isn't which AI model to license or how fast you can ship. It's who you sit down with to find the parts of your business that carry weight and the areas where you can create variance. Choose that partner well, and the rest of the work has a chance of mattering. Choose poorly, and no amount of AI will save you.