The AI Industry Is Rediscovering Something Old

The AI industry talks about data moats as if the concept was invented last year.

It wasn't.

In 1969 David Ness co-founded a company called Interactive Market Systems. Ness was one of the founding architects of the Decision Sciences department at Wharton, a pioneer of Decision Support Systems, and one of the researchers behind Project MAC, the pioneering mainframe computing program at MIT. He was my professor, my mentor, and ultimately my friend.

I learned what follows directly from him.

IMS integrated more than 300 syndicated and proprietary databases. Nielsen. Arbitron. Simmons. MRI. International equivalents across Canada, Europe, Asia, and Latin America. The system helped advertisers answer one question: given a target audience, what is the lowest cost media mix to reach them?

That is programmatic advertising. Built on a mainframe. In 1969.

Ness published the intellectual framework in 1972. The book was titled An Interactive Media Decision Support System. IMS grew into the leading international advertising industry platform with more than 1,000 clients across advertisers, agencies, and media companies on multiple continents.

The principle that built IMS was simple. Proprietary data nobody else had access to. A decision system built on top of it. Competitive advantage that compounded as the data grew.

Ness understood this principle as deeply as anyone alive. In the early 1980s he tutored John Reed on technology before Reed became CEO of Citicorp in 1984. What followed was one of the most aggressive data strategy deployments in banking history. Citicorp deployed ATMs across New York while competitors treated them as a cost center. Every ATM transaction was a data point about customer behavior, spending patterns, and cash flow that competitors simply didn't have.

Same principle. Different industry. Same mind behind both.

The AI industry is not discovering something new. It is rediscovering something old. The organizations that figured this out in 1969 built the infrastructure that defined their industries for decades. The ones that dismissed it fell behind and never fully recovered.

Before your organization asks which AI model to use, ask a harder question. What proprietary data do you have that nobody else can access? What decisions could a system built on that data make faster and more precisely than any competitor?

That is the right first question. It was the right first question in 1969.

History has been teaching this lesson for more than fifty years. The organizations paying attention will win again.