Paco Hope on Nostr: One thing I'm observing at $dayjob with respect to AI is a lot of people automating ...

One thing I'm observing at $dayjob with respect to AI is a lot of people automating processes without asking "what goal does this process achieve?" and "is this a good process for achieving that?" They assume that any process we rely on heavily must be good, so doing that same process faster with AI must be better than doing it without.

A great example is a program that literally launched 3 months ago. The runbooks, policies, documentation are all very incomplete. They're early drafts, they're going to be revised. These policies and runbooks haven't had a lot of exposure to the real world yet, so they don't cover edge cases very well. There are not a lot of "what if's" in them.

But here's somebody emailing us to say they fed the documentation and policies to an agent and isn't it great that the agent can somehow use these policies and runbooks to do something.

I lack the rhetorical grace to diplomatically tell them this was a waste of time. Besides, I'd just be labeled a hater. So I sit quietly and watch the umpteenth half-assed AI app spark and die.
1/2