SAS at 50: The analytics pioneer is cautiously adopting AI

It was the middle of the first AI winter when SAS Institute was incorporated on July 1, 1976, and artificial intelligence was not on its product roadmap. Fifty years on, it’s taking a cautious approach to the technology: not going all-in on AI assistants everywhere, like Microsoft, or all out to build AI infrastructure, like Oracle, but looking for areas where AI can reliably add business value.

What started as a four-person company spun out of a research project at North Carolina State University (NCSU)  is now an analytics and AI giant employing about 11,000 people across 39 countries, with more than $3 billion in annual sales.

And over that half-century the company, still privately held by two of its four founders, has been consistently profitable.

That’s not bad for an organization that had modest ambitions. “When we first formed the company, our goal was to make it through the end of the year and be able to not go broke,” said co-founder and still CEO Dr. Jim Goodnight in an interview. “We actually made a little bit of money that year.”

Development of SAS software began at NCSU in the late 1960s, with the aim to analyze complex agricultural data, and in 1971 it was released to customers outside the university. A 1974 NCSU press release touted it as “a major contribution by North Carolina State University to data analysis in the world,” noting that it was one of the major statistical computing systems in use in the US, and was fast becoming a total analysis system.

After the inaugural SAS user conference in January 1976, Goodnight, along with colleagues A.J. Barr, John Sall, and Jane Helwig, realized that it was impossible to grow further within the university and decided to incorporate. Barr and Helwig both sold their stakes a few years later, leaving Goodnight and Sall as co-owners, which they remain to this day.

From an initial 150 customers using the software when it was still an NCSU project, SAS’s customer base has grown to about 80,000 sites in 150 countries.

The company, headquartered in Cary, North Carolina since 1980, may not be a household name, but its software is behind functions such as data analysis in clinical trials at large pharmaceutical companies, pricing strategy at large retailers, and anti-money laundering and fraud detection efforts in many banks. Over the years, companies in virtually every industry, from aerospace and environmental protection to retail and manufacturing, have used SAS in areas such as data management, risk management, governance, decision intelligence, marketing, and fraud management.

Ease of use

And now, like many other software vendors, SAS is incorporating AI into its offering. Goodnight has a healthy skepticism of some applications of the technology, saying, “people are spending a lot of money guessing the next best word to use in a sentence.” At SAS, the focus is on using AI to make its software easier to use and its answers more self-explanatory in a way that doesn’t leave customers with unexpected bills.

Unlike some other vendors, he said, “When you do a call to AI, it actually comes back to SAS, and we run it here at no charge to the customer, so we tried to add all of our AI capabilities without charging anything for them, because we have the domain expertise.”

The accelerating rate of technology change has been good for SAS, according to its CTO, Bryan Harris. “I think it’s pushed us harder,” he said.

Applying AI with care

While SAS is adopting AI, he said, “we have to be relevant in the hype of a new technology, and most importantly, incredibly relevant in the reality of that technology.” That means spurning things like tokenmaxing, which he called a “vanity metric,” and focusing on business impact and financial responsibility.

“When you start talking about automating business processes in your world with agents, and there is a somewhere between 10% to 30% error rate in those, that is not a good thing, and not something customers can put their careers on,” Harris said. “So, what we show them is how to overcome that error rate, and how we use our technology to do that, and that you need to understand the risks of the technology so you apply it in the right use cases and areas that are appropriate for the business.”

In addition to improving the accuracy and governance of AI agents, Harris said other R&D focuses are on physical AI, digital twins, and quantum computing.

Institutional memory

While the company’s technology has evolved with the times, some things have been consistent, he added. The SAS value system and its people-centric leadership style remain in place, and have repeatedly put the company on the Best Places to Work lists in the US and globally.

“I think we’ve built a great culture,” said Goodnight, now 83. “I hope the ones that carry us forward will remember how to treat people, how to be good to people, and how to pay people well. I hope we see it continue in the future.”