Companies adopting agentic AI prioritize approval ease over capability: Report

A report by the Actuaries Institute reveals that companies adopting agentic AI tools are more likely to choose systems that are easier to approve through cybersecurity and procurement processes, even if those systems are less capable. This trend indicates a potential misalignment between organizational priorities and the actual performance of AI tools. The report underscores that ease of approval may be taking precedence over the functional capabilities of the tools, which could lead to suboptimal outcomes for businesses.

The report notes that the adoption of agentic AI tools is not without its challenges. While these tools are designed to handle high volumes of interactions across various channels, the need for human review of every interaction could undermine their intended purpose. This requirement for oversight may limit the scalability and efficiency of agentic AI systems, which are meant to reduce the burden on human workers.

According to the report, nearly 30 per cent of companies experienced a decline in productivity after implementing agentic AI tools. This suggests that the integration of these tools may not always lead to the expected improvements in efficiency. In one study, AI tools increased coding activity by 180 per cent, but the number of shipped releases only rose by 30 per cent. This discrepancy highlights the potential for AI tools to generate more output without a proportional increase in tangible results.

The consequences of this trend include increased costs due to the need for additional human oversight, potential vendor lock-in as companies become dependent on specific AI platforms, and governance challenges in ensuring that AI systems align with organizational goals. These factors may lead to a fragmented approach to AI adoption, where companies prioritize short-term approval processes over long-term strategic benefits.

As the use of agentic AI continues to develop, the report suggests that organizations must carefully balance the need for approval-friendly tools with the actual capabilities of these systems. The findings indicate that a more nuanced approach to AI adoption is necessary to avoid the pitfalls of prioritizing ease of approval over performance. This ongoing development highlights the importance of continuous evaluation and adaptation in the AI landscape.