SITUATION EXPLAINED: What is the bottleneck for enterprise AI adoption?
We asked @jaminball, partner at @AltimeterCap
"If you think about a large organization, I might be mispermissioned somewhere. I might have access to some system I shouldn't, but as a human, I'm just never even gonna go explore that level of access. I'm not gonna try to access that system that I have access to that I didn't know I have access to because it's not even something I use."
"You think about an agent, they're gonna explore every single nook and cranny. They're gonna find every permission, every piece of context, every database that they are permissioned to have access to."
"Something like a mispermission can have catastrophic effects."
"Big question, concern of enterprises on do I want to adopt agents is how can I govern them, how can I audit them, and how can I know they're doing what they should be doing?"
"How do I get those agents data to help them operate effectively on domain-specific tasks? Those are the two big buckets."
