Seven in ten enterprises cannot identify which AI agent caused a failure in multi-agent environments, according to new research from Kore.ai

New research from Kore.ai reveals that organisations can identify the failures of an AI agent, however, identifying them in a multiagent environment is almost impossible.

In a survey of over 400 IT business leaders, with 2,000 or more employees, 70% said they could identify a failure occurred but could not determine which agent in a multi-agent environment caused it.

Detection time

The time it takes to detect an agent related issue is as low as one to four hours for half of the organisations. However, one third (33%) needed four to eight hours to detect a failure.

Detection measures

The top three metrics for detecting a problem were 39% real-time dashboards, 29% automated threshold alerts, 17% log analysis and pattern detection. However, some 15% of respondents simply relied on end users to report problems as the primary detection metric. Meaning that customers were finding the failure before the organisation.

Lack of trust

On top of issues created by customers finding problems first and reporting it, 53% of organisations reported running agents they don’t fully trust or understand.

Cost of correction

Once teams have gone through all of the above, found a problem, identified it, 79% of respondents said that consequential autonomous actions required a manual reversal and 93% of them said those reversals were costly and disruptive.

“The findings suggest that enterprises are accumulating a form of ‘governance debt,’ where delays in identifying and containing agent failures allow recovery costs to compound long after the initial incident occurs,” said Raj Koneru, CEO and Founder at Kore.ai.

“The market is solving for visibility, but enterprises need accountability. Those are not the same thing. An agent that can be watched but not governed is still a liability. Enterprises do not need better ways to monitor their agents. They need agents built right from the start and governed through production scale,” said Peter Mullen, Chief Marketing Officer, Kore.ai.

To find out more please visit: https://www.kore.ai/research/agent-productivity-index-report