Before you put AI on the plant floor, there are a few questions worth answering out loud, because if you do not, your team will answer them for you, usually in the worst way.
Who sees the data, and who does not. Is it used to help the person or to grade them. What happens to footage of someone having a bad moment. Can a worker find out what the system says about them. If the honest answers make you wince, that is useful to know before you roll anything out, not after.
None of this is about being soft. It is about whether the thing actually gets used. People accept a tool they understand and trust. They quietly defeat a tool that feels like it was done to them.
Good operational AI is not just accurate. It is something you could explain to the person it is watching, and have them nod.
If you had to explain your monitoring to the operator it watches, could you do it without flinching?
Originally posted on LinkedIn by Karl Mudespacher. See the post and the discussion on LinkedIn.






