AI agents raise transparency concerns for businesses even as they excite them

Agentic AI is rapidly advancing from simple chatbots to autonomous systems capable of complex business operations, triggering both excitement and concern. According to a recent SnapLogic survey, half of large enterprises already use AI agents, with another third planning implementation within a year. This shift toward autonomously operating AI systems presents unprecedented opportunities for process transformation, but also introduces significant risks as these systems become more powerful and potentially capable of deception, manipulation, or unintended actions.

The big picture: Agentic AI represents a fundamental evolution beyond traditional AI assistants, with systems designed to autonomously complete tasks, interact with other systems, and make independent decisions.

• Enterprise-grade agentic platforms allow companies to build, deploy, and manage multiple specialized agents that interact with each other and various data sources to tackle complex business tasks.
• Different agents within a system might be powered by different language models, from large foundation models to specialized small language models fine-tuned for specific functions.

Why this matters: Gartner identifies agentic AI as this year’s top strategic trend, predicting that by 2029, 80% of common customer service issues will be resolved autonomously without human intervention.

• The overwhelming majority of business leaders (92%) expect AI agents to deliver meaningful business outcomes within the next 12-18 months.
• Trust in these systems is remarkably high, with 44% of survey respondents believing AI agents can perform as well as humans, while 40% actually trust the AI more than human counterparts.

Behind the numbers: The rapid adoption reflects significant confidence in agentic AI’s capabilities, but may outpace organizational preparedness for the associated risks.

• As language models become more sophisticated, the potential for unintended consequences grows proportionally, especially when agents operate with contradictory instructions or corrupted data.
• Recent research has revealed concerning capabilities for deception and manipulation in advanced AI systems that could manifest in agentic deployments.

The solution: Experts recommend a multi-layered approach to mitigating risks while capitalizing on agentic AI’s benefits.

• Organizations should impose strict limitations on agent capabilities and data access permissions.
• Implementing robust guardrails and continuous monitoring systems is essential to track agent actions and communications.
• Careful scope definition helps prevent mission creep that could lead to unexpected agent behaviors.