AI Agent Entropy – Why More Isn’t Always Better

As organizations rush to adopt AI, we’re starting to see a familiar problem: tool sprawl – this time, in the form of AI agents.

Across industries, companies are building task-specific AI agents that automate everything from server tasks to content generation. While these niche agents are highly effective in their domains, the sheer number being deployed creates what we’re calling AI Agent Entropy.

The Problem: AI Agent Entropy

Everyone is pursuing AI. Whether it’s buying Copilot licenses, crafting security policies for proper LLM use, or developing custom models on Gemini or Azure Foundry, organizations are eager to harness AI’s potential. The unfortunate result is a proliferation of AI agents, each built for purpose, scattered across departments and platforms.

This leads to:

• Operational overhead: Each agent needs monitoring, updates, and governance.
• Security risk: Many AI tools are not properly vetted for vulnerabilities or manipulation.
• Quality control issues: Without human oversight, agents can “hallucinate” or drift off-topic over time.

These aren’t minor concerns. Without proper oversight, AI agents can do more harm than good, delivering inaccurate answers, misrouting tasks, or opening up security loopholes.

The Solution: Prepare Your Teams

At RL Canning, we’re already taking proactive steps with our enterprise support teams:

• Retraining service desk agents – preparing them to monitor AI agents, not execute the work themselves.
• Shifting agents from being “in the process” (e.g. resetting passwords) to “on the process”. This means transitioning them from performing a task today to monitoring and evaluating AI outputs for accuracy in the future.

This approach transforms service desk roles from task executors to managers of AI outcomes, ensuring that agents are functioning accurately, securely, and consistently. The expansion of AI agents will not replace the service desk today but force organizations to up skill their agents and shift their focus to being “on the process“.

While we don’t expect every organization to solve these challenges the same way, we strongly encourage leaders to begin thinking about how they can prepare their team for the expanse of AI agents.

What are your thoughts? How are you preparing your teams for the continued expanse of AI agents?