SAS predicts 2027 will test AI accountability, value and trust

Autonomous AI will raise accountability questions

SAS says the enterprise AI market is approaching a point where organizations will have to answer three questions: who is accountable when an AI system acts, whether it can prove its worth and whether its decisions can be trusted.

As AI agents operate across systems, data and business processes, chief information officers will need to rethink identity, permissions, security controls and monitoring. The infrastructure used to govern human users will need to address autonomous digital systems as well. SAS CIO Jay Upchurch says the agent race will depend on which organizations can trust agents to operate, rather than on which deploy the most.

The company also expects agents to move beyond shopping and purchasing decisions and begin making more consequential financial decisions, including transferring money on behalf of human principals. That will force banks to examine whether a customer authorized an action, whether the authorization covers a particular transaction and whether an agent has been compromised.

For businesses, SAS says responsibility cannot be assigned to the AI itself. Companies will need rules defining what agents can decide independently, what requires human involvement and who has the authority to stop an agent. Unauthorized offers, broken promises or poor customer interactions could become brand crises when customers associate the agent's actions with the company.

Businesses will demand measurable AI returns

SAS predicts an economic reset for AI in 2027, as high valuations, rising debt and the cost of expanding data centers meet pressure to demonstrate sustainable returns. The company says businesses will become more selective and invest where the value justifies the cost.

According to SAS CFO Matt Parsons, hyperscalers could emerge stronger because their infrastructure, customer bases, diversified businesses and scale give them more ways to monetize AI and adapt as the market changes. At the same time, software vendors will need to make purchasing more compelling as AI makes in-house development easier.

SAS identifies domain expertise, product quality and execution as the areas that will distinguish successful software vendors. It also says the economics of AI will increasingly be measured by the cost of delivering trusted business decisions, rather than only by models, tokens and services. Data, governance and infrastructure will become part of the calculation.

The company forecasts a gap between AI productivity and business productivity. Organizations may generate more content, code, analysis and recommendations while failing to improve business performance at the same rate. The constraint will shift toward evaluating, prioritizing and acting on that output. SAS also expects more people to build their own software with AI agents, raising the importance of the enterprise question of whether to build or buy.

Simulation will help test decisions before deployment

SAS expects simulation, digital twins and synthetic data to become increasingly important testing environments for AI. Organizations could use them to examine machinery failures that might endanger factory workers, test procedures in virtual medical environments or stress-test financial systems against unexpected behavior by AI agents.

As agents gain authority to act, simulated environments can give organizations a controlled place to make mistakes, learn and try again before decisions affect the real world. SAS describes this as a way to examine potential outcomes as well as anticipated ones.

The company also predicts that chief data officers and other data leaders will face growing pressure around imperfect decisions as agent deployments increase. They will need to define what makes a good decision, delegate authority, establish acceptable trade-offs and identify key objectives.

Governance will become part of organizational learning

SAS says decision speed will become less important than whether decisions made by people and AI are explainable, ethical and collaborative. Decision intelligence platforms can multiply both good and bad decisions at scale, making the quality of the underlying rules and objectives more consequential.

The company expects governance to support continuous learning by monitoring effectiveness, identifying behavior within or beyond acceptable limits and determining which institutional insights should be retained. Organizations will need to examine not only what AI decided, but also when people intervened and whether those interventions improved the result.

For users and the wider market, SAS's forecast points to a shift from treating AI deployment as the main measure of progress to evaluating the decisions AI helps deliver. The company's predictions cover technology, financial services, insurance, marketing, government, life sciences and healthcare, and frame trust, oversight and demonstrable business value as conditions for broader enterprise adoption.