We tried the zoom AI assistant today at work in a meeting.

The summary it sent to the meeting owner afterwards looks _really quite good_.

Stunning, actually.

BUT

In some places, just factually wrong about what we said. Frustrating. If it doesn't summarize meetings better than 90% then it has a lot less value. Especially in a technical meeting like this where specific decisions/conclusions are made.

It's a great way to get someone who is creating meeting notes/minutes started... but by itself, its almost worse than having no notes at all. I think if we were discussing more generic topics it might have done "better". But on this highly technical discussion with lots of unmentioned context... not so much.

I think this is true of almost all AI applications I've seen recently.

Still, fascinating.