Kore.ai recently published its 2026 Agent Productivity Index, based on a survey of over 400 IT leaders examining enterprise use of AI agents.
According to the report, 72% of enterprises reported that their AI agents operate with unmanaged risks, which respondents associate with potential financial and compliance challenges.
The findings describe AI agents as increasingly involved in handling data, decision-making, and customer interactions. At the same time, the report notes that many IT leaders report limited visibility into how agent-driven decisions are made. This is presented as a concern in cases where agent actions are linked to revenue loss or issues of trust.
Key figures from the report include:
- 79% of enterprises reported having to reverse AI-generated decisions
- 70% experienced issues that were difficult to trace
- 42% reported revenue losses they associated with AI-related failures
The report also states that in 40% of organisations, a single AI agent error affected multiple systems. It further notes that in 2026, 41% of AI agents are reported to be involved in critical operational tasks such as data migrations, system updates, and financial transactions.
Research from organisations such as McKinsey & Company and Gartner is also referenced in relation to broader enterprise AI adoption trends. These sources are commonly cited in discussions about challenges such as governance, costs, and unclear return on investment as organisations scale AI systems, particularly when deployment outpaces supporting oversight frameworks.
The report notes that some enterprises implement additional controls after AI deployment to address emerging risks. It argues that such measures may not resolve underlying design issues, and instead suggests that governance needs to be integrated earlier in system design and operation.
Kore.ai’s Agent Platform (Artemis edition) is described in the report as supporting agent creation from natural language objectives through its AI architecture component, Arch, and includes features intended to support policy enforcement, monitoring, and traceability during deployment and operation across different cloud environments.
The report concludes that enterprise AI agents are associated with higher productivity when governance structures are implemented consistently from the outset, rather than relying primarily on corrective controls after deployment.
