That model made complete sense when every piece of work required a human to initiate it, execute it, and close it. It makes less sense when AI agents can initiate, execute, and close a significant proportion of these workflows autonomously, with humans governing the process rather than running it.

Has anyone here deployed AI agents in a CRE context, even in pilot, that operate with real autonomy rather than just assisted drafting?

What did the governance model look like and what guardrails were non-negotiable before you would let an agent take an action rather than just recommend one?

The floor is yours. Early movers, sceptics, and everyone in between.