SAS urges human oversight in agentic AI deployments

SAS executives have urged companies to keep human judgment at the centre of agentic AI deployments, arguing that stronger returns depend on governance and oversight.

Senior leaders at the software and analytics group said businesses moving beyond pilot projects into AI systems that can act across workflows should avoid treating automation as an end in itself. Organisations are more likely to create durable value when people remain responsible for setting goals, defining limits and making high-risk decisions, they argued.

The comments come as many companies reassess the returns they expect from AI spending. While early projects often focused on cost reduction and task automation, SAS said the more important question is whether businesses can extend automation without removing accountability from the process.

Research cited by SAS from IDC found that organisations that prioritise trustworthy AI are 60% more likely to double return on investment from their AI projects. The company also pointed to findings that customer experience-focused AI initiatives deliver almost 20% higher returns than projects aimed mainly at cutting costs.

Human role

Bryan Harris, Chief Technology Officer at SAS, said human expertise would become more important, not less, as AI tools become widely available.

"When everyone is one prompt away from the same answers, human judgment becomes the differentiator. AI can generate answers, but it cannot replace the people who understand a business, its customers and the decisions that shape its future. The organisations that gain the most from AI will be the ones that use it to amplify human expertise, not replace it," Harris said.

Udo Sglavo, Vice President of Applied AI & Modeling at SAS, echoed that view, saying AI's effect on work will be felt more in changing skills than in replacing entire jobs.

"AI will automate some tasks and change some roles, but the bigger impact will be on the skills people need to succeed. The biggest opportunity isn't turning everyone into a data scientist. It's helping more people make better decisions. Instead of expecting every employee to become an expert, AI can provide expertise when people need it. The companies that get ahead won't be the ones that automate the most work. They'll be the ones that make the best decisions," Sglavo said.

Governance focus

SAS leaders said companies need to build oversight into AI systems from the outset rather than add it at the end of an automated chain. They said this is especially important in situations involving ambiguity, risk or material consequences.

Reggie Townsend, Vice President of Data Ethics at SAS, argued that oversight should sit at the core of system design and use.

"Human oversight cannot be a ceremonial checkpoint at the end of an automated process. If we want AI to reflect human values, people need to be involved in how systems are designed, governed and used. The goal is not simply human in the loop. It is human in the lead - especially when decisions involve ambiguity, risk or real consequences," Townsend said.

The company's position reflects a broader debate in the technology sector over agentic AI, a term used for software systems that can initiate and carry out tasks with less direct supervision than earlier tools. Supporters argue such systems can improve efficiency across workflows, while critics warn they may introduce new operational and compliance risks if staff do not understand how decisions are being made.

ROI debate

For SAS, the case for AI spending rests less on reducing headcount than on improving decision quality and customer outcomes. That shifts the emphasis to where AI is used, and under what controls, rather than to automation rates alone.

Stu Bradley, Senior Vice President of Risk, Fraud and Compliance Solutions at SAS, said returns tend to be stronger in projects tied to customer experience than in those designed mainly to save money.

"Contrary to popular belief, AI's best return isn't headcount reduction. In fact, IDC research commissioned by SAS last year found that customer experience-focused AI initiatives deliver almost 20% higher return than those aimed at cost-cutting. Automation and efficiency matter, yes. But bigger returns come when AI helps humans drive innovation and deliver a better experience, not just a faster one," Bradley said.

SAS also argued that AI literacy is becoming a management issue rather than a narrow technical one. As generative AI tools spread through offices, companies risk employees using unsanctioned systems if internal guidance is weak or inconsistent.

Jay Upchurch, Chief Information Officer at SAS, said senior executives need to explain both the benefits and the risks of AI use inside their organisations.

"AI literacy is now a leadership responsibility. Executives have to become AI translators for their organisations by explaining where AI helps, where it creates risk and how employees should use it responsibly. If leaders do not set clear expectations and model the behaviour themselves, they should not be surprised when shadow AI spreads," Upchurch said.

SAS, which is marking its 50th year, said its long-held approach to analytics and decision-making has informed its position on AI deployment. Businesses that treat human judgment, explainability and accountability as central features of AI programmes are more likely to sustain returns than those focused only on automation, it said.