Software companies used to defend themselves in a handful of reliable ways. They built useful features, made their products easy to use and counted on the hassle of switching platforms to keep their customers from leaving. A new company that wanted to compete had to invest months, sometimes even years, of engineering time to catch up.

AI has demolished that playbook. With today’s tools, a new company can build a full-featured, AI-powered product over a single weekend. That has left nearly every software business facing the same problem. If a competitor can copy what you made almost overnight, the product itself no longer keeps your customers with you.

One of the most durable advantages we see today is not the most advanced AI. It is deep knowledge of a specific industry. The models themselves are available to everyone, but a company that has worked inside one industry for years knows how that industry actually operates, and feeding that knowledge into its AI makes the product far better than any general-purpose tool. That kind of advantage is very difficult to copy, and it favors the kind of company Minnesota and the broader Midwest produce in abundance: software businesses in industries that are complicated, regulated and built on deep domain expertise.

We see this most clearly in vertical software companies. These companies build products to solve a problem for a specific industry. They often become deeply embedded in their customers’ day-to-day work, and those customers naturally turn to them for related needs like payments, insurance and background checks. The result is a core AI-native platform with embedded fintech and other regulated products layered on top. In many cases, those added services drive the majority of revenue. They are the “+” in AI+, a core investment category for Rally Ventures.

That setup is also what makes these companies hard to leave. The more services a platform handles, the more a customer would have to unwind to switch because payments, insurance and other products are now woven into daily operations. And the longer a platform runs inside an industry, the more it understands how that industry actually works, from daily workflows to the patterns that only show up over time. That accumulated knowledge is the industry context behind the better AI, and it shows up in four different parts:

The first part is operational history. A software platform that sits at the center of how a business runs collects years of information, like which tasks get done in what order, which decisions get reversed and where the delays pile up. When a company adds AI on top of that record, the AI has experience to learn from. A competitor building the same feature from a general-purpose model is starting with none of that context.

The second part is data that exists nowhere else. Take construction as an example. A general AI tool can offer a rough estimate of what a project should cost. A construction software platform that has tracked every material order and every budget overrun across thousands of projects can produce an estimate that is far more accurate, because it is drawing on information no outsider has. That kind of data cannot be purchased or scraped from the internet. A company earns it only by working inside an industry for years.

The third part is connection to the older systems that businesses actually run on. Most company data does not live in modern software. It lives in banking systems, insurance claims platforms and medical records software installed long ago. A company that has already done the difficult work of connecting to those systems can feed live information into its AI. The human security platform Yardstik spent years wiring itself into the scattered sources a background check depends on, from FBI and motor vehicle records to fingerprinting systems and courthouse databases in counties across the country. That work is now the foundation its AI runs on, and it is not work a newcomer can skip.

The fourth part is trust, which I would argue matters most of all. A hospital or an insurance company will not hand sensitive data to a startup that launched a few months ago. It shares that information with a platform it has come to rely on over years. The security audits, licenses and partnerships required to earn that access take a long time to build, which is what makes them so hard to reproduce.

These four things reinforce one another, and they compound. More customers mean more data, which makes the AI more accurate, which keeps customers longer and brings in new ones. The models these companies use are largely the same ones available to everyone. The data feeding them is not, and it grows richer every year.

The national conversation about AI tends to focus on consumer apps and the companies building the underlying models. The winners that get less attention are the specialized software companies working behind the scenes in healthcare, insurance, agriculture and financial services: industries Minnesota has built companies around for decades. These companies generally aren’t very flashy, but what they have is exactly what the AI era rewards: a deep, hard-won understanding of how a particular industry actually works.

Having the best technology was never going to be enough, because eventually everyone has access to the same technology. What a competitor cannot download is the decades of deeply knowing an industry, and a great many of the companies that hold that type of knowledge are right here in Minnesota.

Justin Kaufenberg is the Managing Director at Rally Ventures.