SAS CEO Dr. Jim Goodnight and œżÉ«”ŒșœJay Upchurch share perspectives on AI, customer requests, and the companyâs upcoming IPO.
SAS Institute cofounder and CEO Dr. Jim Goodnight has been steeping his company in artificial intelligence for decades â long before it became cool.
âAI, which I consider to be neural networks, machine learning, that is really now being called âclassical AI,ââ Dr. Goodnight told CIO.com during the companyâs recent SAS Innovate conference. âThatâs what weâve been doing; weâve been doing neural networks for over 20 years.â
The latest AI wave â generative AI â is something completely different, he remarked.
Although gen AI, too, uses neural networks, it uses a vector âto put you in the right area to start answering the question,â he explained. From there it computes which word has the highest probability of coming next.
âAnd thatâs all it is,â Dr. Goodnight quipped. âSo, people are spending millions and millions of dollars trying to figure out whatâs the next word in a sentence.â
The multinational SAS is considered the worldâs largest privately held software company. In 2021, it , a deal reported to have been . The companyâs âclassical AIâ and analytics offerings have underpinned its customersâ efforts to become more data-driven for years, and its cloud-native Viya platform, originally released in 2016, , complete with AI agents, copilots, and synthetic data tools, as well as with AI governance baked in.
CIO.com sat down with and SAS during SAS Innovate for wide-ranging discussions about the companyâs direction, the state of AI use today, and the software makerâs long talked-about upcoming IPO.
AI gets agentic
Even as Dr. Goodnight differentiates generative AI from SASâs legacy in machine learning AI, he still sees significant value in its potential. SAS does use gen AI to assist with code generation, enabling users to ask its assistant a question in natural language and the AI will generate the SAS code that answers it. And, he said, the code is pretty good â on a level of âthe first program you ever wrote.â
However, he thinks companies by and large are just telling staff to âgo do this gen AI stuff,â and many, he believes, are wasting a tremendous amount of money feeding such systems company-specific training data.
Call centers and language translation are proving to be top use cases for gen AI, âso weâve done a lot of work with it,â he said, but âitâs not something I think is going to be the future of the world, or anything like that.â
Agentic AI, which focuses on operational decision-making, is AIâs latest incarnation, but the lack of a standard description of what makes a true AI agent is creating confusion in the technologyâs early days. âItâs just the latest buzzword to talk about programs that help you make a decision,â Dr. Goodnight said, noting that an agent can just be âa bunch of âifâ statements and logic.â
âThe thermostat on your wall is an agentic AI,â he explained. âIt is a mechanism that senses and has sensors to sense the temperature. It has logic to know when to turn the air conditioner off, and it has a mechanism and a capability of doing that. So your thermostat is a probably an agentic AI thatâs been around for a number of years.â
Still, advances in the technology are having an impact. SAS œżÉ«”ŒșœUpchurch noted that decisions powered by the companyâs new SAS Intelligent Decisioning are proving popular with preview customers.
âI think decisioning, especially our intelligent decision capability, is really the next forefront of bringing agentic AI to life for enterprise,â he said. âWhen we were talking to Microsoft about it over a year ago, I think the lightbulb went off in their mind around, âWow, Iâve got to have that embedded in Fabric.â We brought that to life now for them.â
What SAS customers are asking for
Another area where SAS sees increasing customer interest is around synthetic data â computer-generated data that can help companies with AI training challenges without the customer privacy concerns.
When late last year, customers didnât know much about the SAS foray into synthetic data, Upchurch said.
âI think we brought it to life more this week than we have in the past,â he said of the companyâs announcements around its synthetic data strategy at SAS Innovate. Now customers are seeing real-world applications. He spoke to one whose data sovereignty problems meant that they had to fly people in a central location to work on sensitive data who now plans to create a derivative synthetic dataset that they can move around the world without legal issues.
Upchurch added: âThatâs real-world applicability that they can take home from the conference and immediately show their partners that theyâve got a solution and a trusted partner, SAS, to help them with it.â
Customers are also asking for help with their large language models (LLMs), said Upchurch, noting that SAS can be âa trusted partner in that space.â
âWe want to make large language models better, or we want them for use embedded in our solutions,â he said. âItâs not exactly where weâve invested so far, [but] I think thatâs going to continue to evolve as people move further away from just an LLM and more into: Whatâs the real-world applicability for AI and agentic AI, and how do you make it real for your employees?â
Upchurch also sees customer appeal in the companyâs AI governance initiative, which began in 2021 as a passion project led by Reggie Townsend, now VP of data ethics at SAS. Dr. Goodnight has been âso supportive, and Iâll say gracious, with investment,â Upchurch said, adding that the products SAS has created, such as its newly announced AI Governance Map, are based on customer demand.
And âyouâre seeing us turn around and give it away for free. I know there are plenty of others out there that are trying to make it much more of a commercial offering,â Upchurch said. The SAS offerings âare available to anybody. I think thatâs indicative of our belief in the importance of this. And I think itâs a testament to Dr. Goodnightâs belief, obviously, and what weâre trying to do around trust and transparency.â
The promise of customer choice
Although SAS, like most other vendors, is leaning toward the cloud, the company continues to keep Dr. Goodnightâs promise from many years ago that customers would always have an on-premises option, Upchurch said.
âWe definitely, as a corporate strategy, want to empower customer choice,â he said. âIf you want to run it on prem, you can; if you want to run it in your cloud of choice, you can. And if you want SAS to run and manage it for you in your cloud choice, we can.â
The SAS 9 (on prem) engine is âgreatly valuedâ by customers; the company continues to maintain it, and thereâs a major release in the works, Upchurch said.
That said, SAS is actively encouraging customers to move to Viya, where the company is focusing its innovation, he said. And it seems to be working.
âI have seen more customers in the last 12 to 18 months move directly from an on-prem version of SAS 9 to a hosted version of Viya than I had probably seen in the previous couple years combined,â he observed. âOftentimes, customers are wanting us to lift SAS 9 from on premises, put it in the cloud and then run Viya side by side. And now what weâre seeing is the adoption is gaining speed, and itâs getting easier for our customers to move workloads from SAS 9 to Viya.â
The long, long journey to IPO
Earlier this month, the Cary, N.C.-based company announced that SAS 13-year veteran with a remit in part to help prepare the company for an IPO. Â
Itâs not the first time going public has been on the table for SAS.
Since the notion of a SAS IPO was , the date has been pushed farther into the future each year. The company, which will celebrate its 50th anniversary in 2026, first made the announcement at a user conference in 2000. That didnât happen after the stock market collapsed.
In 2021, , with the company saying an IPO was planned for 2024.
There were, however, challenges, Dr. Goodnight told CIO.com.
For example, the company had to consolidate 14 financial systems, which took several years; that new system went live in January 2025. SAS also had to build other systems necessary for public companies, and those went into production at the beginning of May.
âWeâll need to run on that for at least a year before weâre ready,â he said, so the IPO may happen next year, although itâs more likely to occur in 2027.
