Menu
SUBSCRIBE
SUBSCRIBE
Menu

enterprise AI adoption

A white globe with continents made of blue dots, from the cover of a SAS and IDC report on trust in agentic AI

IDC Report Finds Businesses Give AI Agents Decision Roles As Trust Lags

, , , , , , , , , ,

Nearly 9 in 10 organizations said AI agents play some role in decisions, but only 66% of respondents said they trust agentic AI.

 

Key Takeaways

  • An IDC survey found that 89% of organizations give AI agents some role in decisions, while 66% of respondents trust agentic AI.
  • 79% of respondents said people override AI recommendations in more than 10% of cases, most often because the result lacks an explanation or important context.
  • Only 17.5% said their organizations had fully optimized their data systems.

A new SAS report from IDC found that companies are allowing AI agents to influence or make decisions faster than trust in those systems is developing.

The report is based on an April 2026 IDC survey of 2,699 decision-makers across 28 countries. It focused on banking, insurance, life sciences, and the public sector.

Decision-making authority expands faster than trust

Survey respondents said AI agents play a role in decisions at 89% of their organizations, from recommending actions for people to approve to acting without prior approval.

Only 66% said they trust agentic AI, leaving a 23-percentage-point gap between reported use and reported trust. Agentic AI also trailed generative AI, which 76% said they trust.

Users intervene when AI cannot explain its work

The survey found that 79% of respondents override AI recommendations in more than 10% of cases.

The leading reason was insufficient explanation, cited by 34.5%. Another 25.6% said an AI result failed to account for the full context of the decision. Factual errors ranked lower, at 14.8%.

Data systems remain a governance weakness

Only 17.5% of respondents said their organizations had fully optimized their data systems. The report said mature data systems make it easier to trace information to its source and continually check whether an AI system is producing reliable results.

Stronger governance is associated with higher reported returns

Organizations with the strongest AI governance practices were far more likely to report high returns on their AI spending. The report gave higher scores to organizations with stronger controls over the data AI uses, how it reviews decisions, and who is accountable for its use.

Among the highest-scoring organizations, 62% reported earning at least $2 for every $1 spent on AI, compared with 4% of the lowest-scoring organizations.

The survey does not establish that stronger governance caused the higher returns.

The report said 86% of organizations plan to increase spending on AI oversight, data quality, validation, and accountability during the next year. It did not specify how much each organization plans to spend.

Illustration of a human head in blue neural network lines beside an AI chip, symbolizing enterprise AI governance.

AI Governance Continues To Trail Enterprise AI Adoption, ISACA Survey Finds

, , , , , , , , , , , ,

The global survey found AI is now a standard part of business operations, but AI governance continues to lag behind the pace of adoption.

Key Takeaways

  • 90% of respondents said that employees in their organizations use AI.
  • Only 38% said their organization has a formal, comprehensive AI policy.
  • 45% said AI risk is an immediate organizational priority.
  • Only 38% expressed confidence in their board’s understanding of and action against AI risks.
  • Just 12% said their organization has a documented AI shutdown process they test regularly.

AI has become a standard part of enterprise operations, but governance frameworks continue to lag behind adoption, according to ISACA’s 2026 Taking the Pulse of AI survey.

90% of respondents reported that employees are using AI within their organization, while 81% said employees are using generative AI specifically. Based on responses from more than 3,400 technology, governance, risk, and cybersecurity professionals worldwide, the findings suggest AI has moved from experimentation to everyday business use.

Governance lags behind AI adoption

Despite AI use being commonplace, many organizations have yet to establish guardrails regarding its use.

Only 38% of organizations reported having a formal, comprehensive AI policy, up from 28% in 2025. Another 30% said they have a limited policy, while one-quarter said their organization has no active AI policy at all. Meanwhile, just 45% of respondents said AI risk is an immediate organizational priority, indicating that governance efforts continue to lag behind AI deployment.

The survey also found that AI literacy is becoming an essential workplace skill. 78% of respondents said AI skills are now very or extremely important to their profession, up from 72% a year earlier. At the same time, only one-third of organizations now provide AI training to all employees.

Board oversight still has gaps

The governance challenge extends into the boardroom.

Only 38% of respondents said they are confident their board understands AI risks and is taking appropriate action. The finding suggests many organizations have not yet fully incorporated AI governance into executive oversight, even as AI becomes embedded across business operations.

The survey found that respondents remain most concerned about AI’s potential to create security and operational risks, even as organizations continue to expand their use of the technology. However, only 36% expressed confidence in their organization’s ability to detect AI-generated misinformation.

Organizations aren’t ready to respond

The survey also points to significant gaps in operational preparedness should an AI system malfunction or be compromised.

Only 12% of respondents said their organization has a documented AI shutdown or override process that is regularly tested. Another 8% said a documented process exists but has not been tested, while 39% did not know whether such a process exists at all. Separately, 56% said they do not know how long it would take their organization to halt an AI system following a security incident.

 

The report is the latest to demonstrate a gap between AI adaptation and governance. In recent months, AI Risk Today published results of several reports centered on this “governance gap.” All are pointing to the same problem: with AI adoption accelerating, governance planning is still well behind.