Honestly, the biggest 'gotcha' for me was realizing that even in a fully automated factory, you can't escape the human element completely. I thought we'd just set up the robots and AI systems and let them run 24/7, but here's what really happened:

The maintenance requirements were way more intense than expected. You'd think automation means less maintenance, but actually, it's just different - and often more complex. When a robot arm fails at 2 AM, there's no operator nearby to notice subtle signs of trouble. The systems need constant monitoring and predictive maintenance, which requires specialized skills that are hard to find.

Software integration was definitely a nightmare too. Getting all these different systems - robotics, AI, sensors, inventory management - to actually talk to each other smoothly was like herding cats. We'd have one system working perfectly, then add another and suddenly nothing would communicate properly.

But the real eye-opener was discovering that some processes just need human judgment. Quality control is a great example - AI can spot obvious defects, but subtle variations in materials or complex assembly issues still need that human touch. We also found that unexpected situations (like a part that's slightly out of spec but still usable) require decisions that our automated systems couldn't make.

The most frustrating part? Thinking you're building a 'lights-out' factory, only to realize you're actually building a 'dimly-lit' factory that still needs human oversight, just in different ways and at different times.