Should AI Agents Be Banned From Board Governance?

My colleague Myra McGovern recently shared a question that a school board debated while updating its bylaws. I was struck by how thoughtful and forward-looking the conversation was.

"Should we explicitly state that a trustee must participate personally in board discussions and votes, rather than delegating that role to an AI agent?"

As a school leader, does that question sound absurd? Too science fiction to worry about right now?

The more I thought about it, the more interesting the question became.

I can imagine a highly technical trustee asking an AI system to "keep an eye on board conversations and brief me on anything important." That is completely doable today. And once AI is monitoring a workflow, it is not a huge leap to imagine it drafting responses or recommendations. Some emerging systems are already moving toward agent models that watch workflows, flag decisions, and propose actions.

A trustee might absolutely use AI to prepare, analyze documents, or test questions ahead of a meeting. In that sense, AI could strengthen governance.

But governance itself is fundamentally human work. Trustees are not only analyzing information. They are exercising judgment, accountability, and stewardship of mission. Governance decisions are exactly the kind of responsibility that should remain firmly human in the loop.

I doubt many people would argue with that. The real risk isn't that AI replaces the trustee — it's that a well-intentioned technical trustee deploys an AI system without the right guardrails.

I bring this up because many people will hear this scenario and assume it is unrealistic. In reality, the underlying technology is advancing quickly enough that situations like this will begin to appear in unexpected places.

Whether this belongs in legal policy is a question for attorneys and risk committees.

But for leaders, it may be worth recognizing that the boundary between human participation and AI assistance will become more complex in the years ahead.

And boards may eventually have to decide where that boundary belongs.

Previous
Previous

AI Is Starting to Understand Intent

Next
Next

"It's Faster to Just Do It Myself" — The Efficiency Paradox