BS, But Not Forever

I saw an announcement of more agentic AI stuff, this time from ServiceNow. I don't even feel like listing out details, because if you unfocus your eyes you can just pretend this is a press release from any of one hundred billion enterprise software companies. Here's a little caution, with an AI Kool-Aid chaser at the end.

• Agents, as they are right now, should not be viewed as people-replacers. In their current best form, they are task-replacers. To best take advantage of them, you need to deeply understand how to measure and manage the things the agents will perform.

• Take every agent claim with a huge grain of salt. Do NOT just watch demo videos - insist on trials and stress-testing agents. Know what you expect and need, and insist on that. A good rule of thumb would be to look at the lofty promises and expect only 60%-70% of them to work on your first trip.

Now for the Kool-Aid: I am probably going to eat these words in very short order. If you are paying attention to the crazy advances (and the speed of those advances), you'll see that AI of various flavors continues to quickly climb the charts of any measures we dream up. I have little doubt that what is BS for agents now (specifically, any hype that you can fire-and-forget them at complex tasks) will graduate to non-BS soon. PM