%201.webp)
SAS has published its predictions for how artificial intelligence will affect business in 2027, arguing that accountability, trust and proof of value will be the main tests of corporate AI use.
The forecasts point to a shift from experimental AI projects to broader operational deployment, especially as AI agents take on more autonomous tasks across finance, customer service, software development and decision-making. SAS says companies will face tougher questions about who controls these systems, how their decisions are checked and whether AI spending delivers measurable returns.
Autonomy and control
One theme is governance as AI systems move from recommending actions to taking them. For Chief Information Officers, SAS argues, the issue will no longer be limited to which employees can access systems, but what autonomous agents can do inside them.
"The charge of the CIO in 2027 won't just be deciding who has access to enterprise systems, but what does. With AI agents at work across systems, data and business processes, IT will have to reimagine enterprise infrastructure that was built to govern human users, not autonomous digital ones. In the new year, CIOs will rethink identity, permissions, security controls and monitoring accordingly. What can an agent access? What is it authorised to do? Who is accountable when it acts? CIOs who can answer those questions without burying agents in friction will be best positioned to deploy them. The agent race won't be won by who deploys the most; it'll be won by who can trust them to operate," said Jay Upchurch, Chief Information Officer, SAS.
SAS expects the same issue to extend into banking and payments as AI agents begin making more significant financial decisions on behalf of customers. That would force banks to look beyond transaction monitoring and consider whether an agent has valid authority and whether it has been compromised.
"In 2027, AI agents will graduate from making shopping and purchasing decisions to making more consequential financial decisions, like transferring money on behalf of their human principals. That opens Pandora's box, complicating fundamental questions of authority: Did the customer authorise the agent? Does that authorisation extend to this transaction? Is the agent acting within its mandate, or has it been compromised? As machines gain greater autonomy over money movement, banks will be forced to redefine the boundaries of fraud prevention. Beyond verifying individual transactions, they'll need to ascertain the scope and integrity of the agents behind them," said Adam Neiberg, Global Banking Manager, SAS.
SAS also predicts that ownership of AI actions will become a board-level and regulatory issue. Companies, it argues, will not be able to distance themselves from errors or harmful outcomes created by an agent in a customer-facing role.
"In 2027, 'Who owns this AI agent?' will become as fundamental a governance question as 'Who owns this budget?' Companies that cannot answer it will eventually have to answer to their customers, boards or regulators. As AI agents move from recommending actions to taking them, the risk changes. If your AI says it, your brand says it. An unauthorised offer, a broken promise or a bad customer interaction can quickly become a brand crisis. You can't delegate accountability. Companies will need clear rules for what an agent can decide independently, what needs human involvement and who has the authority to stop it," said Jenn Chase, Chief Marketing Officer, SAS.
Value under scrutiny
A second theme is economics. SAS expects businesses to scrutinise AI spending more closely as heavy investment in infrastructure and high market valuations come under pressure to produce financial returns. In that environment, larger cloud providers are likely to prove more resilient because they have broader revenue bases and greater flexibility.
"In 2027, the economics behind the AI boom will face a major reset as sky-high valuations, mounting debt and the enormous cost of data centre expansion collide with pressure to prove sustainable returns. FOMO helped fuel unprecedented investment, and growing scrutiny of practices like round-tripping will put even more focus on whether the math holds up. AI will continue to advance, but businesses will become far more selective, investing where the value justifies the cost. Hyperscalers will emerge stronger because their diversified businesses, infrastructure, customer bases and scale give them more ways to monetise AI and the financial flexibility to adapt as the market resets," said Matt Parsons, Chief Financial Officer, SAS.
Software suppliers will also face pressure to show why customers should buy products rather than build their own with AI tools. SAS predicts that easier software creation will reshape procurement decisions and could weigh on subscription models as more users develop tailored applications themselves.
SAS links that trend to a broader reassessment of productivity. Rising output in the form of code, content and analysis, it says, will not automatically translate into stronger business performance if organisations fail to redesign workflows and preserve human judgement.
Testing decisions
The third theme is decision quality. SAS predicts greater use of simulation, digital twins and synthetic data to test AI systems before they affect real-world operations, from factory safety to healthcare procedures and financial systems.
It also expects senior data leaders to focus less on decision speed and more on whether machine-led decisions are explainable, ethical and collaborative. That would make governance part of an ongoing learning loop rather than a compliance exercise.
"The conversation will evolve from risk management and controls to learning and innovation. The better an organisation learns, the better it can innovate. As agents act faster and more decisions are automated, organisations will need to understand not just what AI decided, but when people intervened and whether those interventions improved the outcome. Governance will be core to the continuous learning process: monitoring and guiding effectiveness, identifying preferred behaviours and those outside acceptable limits, and determining which institutional insights are worth retaining. In an era of commoditised intelligence, the organisations that learn best will be best positioned for competitive advantage," said Reggie Townsend, Vice President, AI Ethics, Governance and Social Impact, SAS.