Who Are AI’s Horses?

The automobile changed the world. It didn’t make every automobile investor rich.

At the beginning of the twentieth century, investors could have predicted that automobiles would transform civilization. And those investors would have been right.

But knowing that automobiles would succeed was not the same as knowing which automobile manufacturers would make money. Hundreds of companies entered the business, most disappeared.

The horses were in trouble.

You didn’t need to know whether Ford, General Motors, or some forgotten manufacturer would win. You only needed to recognize that a machine capable of doing a horse’s job faster, cheaper, and more reliably was going to make life difficult for the horse business.

Artificial intelligence presents a similar puzzle.

Investors are spending considerable energy trying to identify the next great AI winner. Chipmakers, model developers, cloud providers, and software companies are all competing for a prize whose eventual distribution remains uncertain.

Some could become enormously valuable businesses. Others will discover that creating tremendous value for customers and capturing that value for shareholders are two very different things.

The more interesting question may be a simpler one.

Who are the horses?

Consider a business whose principal economic advantage is employing thousands of people to perform repetitive intellectual work. Perhaps they answer customer inquiries, process insurance claims, write routine software, prepare standard reports, or move information between systems.

For decades, such businesses have enjoyed a perfectly sensible economic model: hire people at one cost, sell their labor at a higher price, and pocket the difference.

The model works so long as the customer needs the people.

But what happens when the customer needs only the result?

If AI can perform a meaningful share of that work at a fraction of the cost, the economics begin to change. And unlike the automobile, which required roads, factories, fuel stations, and decades of physical replacement, software can spread through an existing digital infrastructure remarkably quickly.

That doesn’t mean every outsourcing company, consultancy, or software vendor is doomed.

Some own deep customer relationships, proprietary data, regulatory expertise, distribution, or systems that are genuinely difficult to replace. Those advantages can survive technological change. Indeed, they may become more valuable when combined with AI.

But a company whose principal moat is the number of hours it can bill has a problem when the number of hours required begins to collapse.

The same question applies to software businesses that charge substantial fees for relatively simple workflows. If the customer can accomplish the same task through an intelligent assistant, the old price may become difficult to defend.

There is an important distinction between identifying a threatened business and making money betting against its stock.

Horses didn’t disappear the day the first automobile rolled onto the road. They remained useful for years. Businesses adapt, customers resist change, regulations intervene, and investors can lose fortunes being right too early.

Nor should we assume that the companies supplying the new technology will necessarily capture most of the profits. The history of aviation offers another useful warning: an invention can transform human life while delivering disappointing returns to generations of investors.

The central investment question, then, is not whether artificial intelligence will change the world. That is already happening. It is who will retain the power to charge for what they do once intelligence itself becomes cheaper and more abundant.

Investors naturally gravitate toward the shiny new thing. It’s exciting, it makes impressive demonstrations, and everyone wants to own a piece of the future.

But sometimes the better investment insight is found by looking in the other direction.

Not at the automobile.

At the stable.