So Ive been a PM for about eight years now mainly doing standard SaaS stuff and mobile apps and I thought I had my process down to a science but man this new AI project is kicking my teeth in. Im currently based in Austin working for this mid-sized industrial firm and we have a $200k budget to build a predictive maintenance model for their assembly lines over the next six months. Standard enough right? Wrong. I realized about three weeks in that my usual toolkit isnt quite cutting it because the devs are talking about stochastic gradients and data drift while the CEO is asking why the robot isnt 100% accurate yet. It feels like Im constantly playing translator but in a way that involves managing a lot more uncertainty than I ever had with regular code.

With regular software if it breaks you fix the logic but with this model its like... we have to wait and see if the data even supports the outcome we want. It's making me realize that maybe the soft skills I used for years arent exactly what I need here. Like how do you tell a stakeholder that the timeline is - it depends on the training results - without sounding incompetent? I feel like I need to develop a very specific type of patience or maybe a better way to explain probabilistic outcomes to people who want binary answers.

I was talking to one of our data scientists and she was getting frustrated because I was pushing for a hard feature freeze and she was like look the model needs more tuning. It was a wake up call. What are the actually essential soft skills you guys have found for managing these types of projects? Is it more about managing the ego of the researchers or is it about being a shield for the team when the output looks like garbage for the first two months? Id love to hear what actually works when you're in the trenches because my usual organizer vibe is struggling with the chaos of machine learning...