AI agent accountability is turning into a real worry for IT leaders. A new survey released by Workato on 13 October 2026 found that 74% of US IT leaders expect IT leadership to be held accountable if an AI agent breaks policy, exposes data or harms a customer. Many of the same leaders admit they don’t fully control those agents.
The report, titled AI Challenges Threatening IT Leader Success, is based on 1,001 US IT decision-makers at VP level and above, at organisations with more than 1,000 employees.
The AI agent accountability gap
According to Workato, the findings show leaders carrying responsibility despite gaps in visibility and oversight:
- 74% said IT leadership would be held accountable for a harmful agent action, more than twice as many as those who said the AI vendor would be.
- 23% said they would personally be expected to resign or be removed after a major public AI incident.
- 62% agree that most AI agents in use at their organisation were not formally approved by IT.
- 60% worry more about AI agents than traditional cyberattacks.
- 63% said their IT team knows less about how AI agents work than other employees.
“IT leaders are in a double bind. They are pushed to roll out increasingly powerful AI, yet they can’t fully control it, and they are on the hook when it goes wrong,” said Adam Seligman, CTO of Workato.
The security concern: hiding incidents
The most worrying numbers are about what happens after something goes wrong. Nearly two-thirds (62%) would bet an AI agent will cause a public security or compliance incident at their own company within 18 months. If a major incident happened tomorrow, 36% said their organisation would try to handle it quietly, without disclosure. Nearly half (46%) said the board would demand an external audit, and 24% said their company would publicly blame the AI vendor.
Shadow AI adds to the risk. Workato says about a third of agent adoption is deployed directly by business units and employees without oversight, and 64% of IT leaders expect unauthorised deployments to accelerate. Agents that nobody approved, acting on company data, are exactly the kind of blind spot attackers look for. Our earlier story on how AI attacks add $1M to breach costs shows what is at stake.
Leaders still want more AI
Despite the fears, nobody is pulling back. More than 95% believe the benefits of agentic AI and “digital coworkers” are overwhelmingly positive, and 92% expect their organisations to adopt digital coworkers in the next 18 months.
What they want is control. Visibility into what agents are actually doing ranked first at 42%, rising to 48% among CTOs. That was followed by the ability to audit and reverse an agent action (31%), knowing exactly who is accountable (29%) and confidence that every action is compliant by default (27%). Nearly half are changing their integration and orchestration strategy, and 45% have added AI agents to board-level discussions.
Read the numbers with care
There is another side to this. Workato sells a platform for controlling and running enterprise AI, so a survey showing that leaders need more control supports its own business. The research was conducted by Censuswide among US respondents between 15 and 22 May 2026, so it may not reflect other markets, including India. Many of the figures are expectations and bets, not records of real incidents.
Even so, the message matches what we see as companies rush into agents, from the HCLTech and Anthropic partnership to India’s ongoing AI regulation consultation.
My view on AI agent accountability
As a writer, I feel the 36% figure is the one to remember. Mistakes by AI agents are probably unavoidable, but quietly covering them up is a choice. If companies want people to trust agents with their data, clear ownership, audit trails and honest disclosure have to come before scale. I’d also like to see similar data from Indian firms, where agent adoption is speeding up.
FAQ: AI agent accountability survey
Who ran the survey?
Censuswide conducted it for Workato among 1,001 US IT decision-makers at VP level and above, at companies with 1,000+ employees, between 15 and 22 May 2026.
What is the main finding?
74% said IT leadership would be held accountable if an AI agent violated policy, exposed data or harmed a customer.
What do IT leaders want most?
Visibility into what AI agents are doing across systems ranked first at 42%, followed by the ability to audit and reverse agent actions.