Enterprise AI is at a pivotal juncture the place CIOs should gauge the automation and productiveness advantages AI can convey whereas demonstrating how AI delivers ROI and enterprise worth to their organizations. A elementary requirement of the CIO job in 2026 shall be to find out when launching new AI know-how stops being additive to a company’s aim of creating environment friendly workflows and lowering prices, defined Jay Upchurch, CIO for BI firm SAS .
That is the 12 months through which CIOs will transition into the function of “Chief Integration Officers,” figuring out the way to combine current AI applied sciences inside their organizations as an alternative of frequently launching new AI techniques, predicts Upchurch.
“The CIO function, particularly in ’26, goes to need to be rather more rooted in integration — integration between [AI] brokers and techniques … after which ensuring that we’re doing that in a safe and ruled method,” Upchurch mentioned.
(Supply: Shane Snider/Knowledge Middle Information) SAS CIO Jay Upchurch in a dialog with InformationWeek’s Kelsey Ziser.
Over the previous two years, CIOs have been pressured to implement Gen AI — and later agentic AI — as quick as attainable, mentioned Upchurch. Nevertheless, demonstrating ROI from launching these AI applied sciences hasn’t been straightforward, making a “cost-efficiency dilemma for all CIOs,” he mentioned.
He likens the problem in translating AI deployments into ROI to the early days of cloud computing and virtualization. In the course of the cloud transition, CIOs “needed to refactor all our workloads and reimagine how we ran with that,” he mentioned. “It is the identical method with AI — AI for automation’s sake is just not sufficient.”
The main focus now ought to be on the way to combine current AI investments.
Managing the AI integration course of
Basis piece. The method for integrating AI applied sciences begins with figuring out the “agentic basis” that a company goes to make use of, based on Upchurch. For instance, SAS is a “Microsoft store,” he mentioned, so it is utilizing Microsoft Copilot because the foundational piece to make sure compatibility throughout current productiveness instruments. “We want that digital assistant that may be a pure to our circulate of labor.”
No matter which vendor a CIO selects for the group’s important AI platform, it is necessary to have a basis that works throughout a number of IT techniques, he underscored.
Integration choices: Construct vs. purchase. As soon as a CIO has chosen the foundational AI system, they will get to work on integrating that foundational AI with different AI applied sciences that the group has deployed. However CIOs ought to remember that this course of will take time as a result of AI applied sciences are nonetheless evolving, Upchurch mentioned.
“I’ll have an enterprise system that does not have an agent but that I can talk with,” he mentioned. “That does not imply I ought to rush and construct one. I ought to be affected person sufficient that I pay cash to that enterprise software program vendor to allow them to present up with the agent that I can then combine in with. I haven’t got to construct all of it myself.”
Safety and governance. The third part that CIOs ought to give attention to through the AI integration course of is safety, knowledge and entry, Upchurch mentioned. CIOs might want to make sure that workers have entry to solely knowledge that is related to their function.
“So when that agentic integration is going on, it is figuring out what roles are, what [employees] are entitled to know from a knowledge standpoint — ensuring that label on all knowledge and the opposite techniques is constant so you are not seeing belongings you should not,” he mentioned.
