Agentic AI Requires Oversight and Governance in Legal Departments

In a LawNext article on OpenAI’s new ChatGPT Agent, Ken Crutchfield writes that agentic artificial intelligence requires oversight and governance. Agentic AI refers to autonomous agents that can reason, make decisions, and direct other systems. While this capability offers vast potential, it also carries significant risks, particularly in task mismanagement and cybersecurity. 

Crutchfield notes that many experts prefer the term “semi-autonomous” rather than autonomous to emphasize the need for human supervision and guardrails. The technology demands careful integration within legal departments, with controlled access to services and thorough training for those using or benefiting from agents.  

Because it may be prone to errors in basic tasks, agentic AI requires oversight and governance. Agent-driven workflows must include human oversight, while AI governance frameworks should be expanded to address agent behavior and permissions. Before it can act, the ChatGPT Agent must have explicit user permission. Even seemingly obvious instructions may need to be spelled out.

Clear processes, defined roles, and established escalation paths are essential for effective agentic AI management. Accountability measures need to be in place to maintain trust and compliance. The focus should remain on achieving meaningful outcomes while making use of existing processes. Untested AI features may be outperformed by proven systems.

As AI technology has evolved from a reactive assistant to a proactive task management tool, the release of the ChatGPT Agent has underscored the need for solutions that are rooted in people, processes, and prudent supervision.