Optimization, not just automation
Automation makes a process run on its own. Optimization makes it run better. The difference is most of the value — and most of the skill.
Automating a process is, in a sense, the easy part. You capture what an operator does, you encode it, and the machine does it consistently. That's worth a lot. But it's also where most projects stop — and stopping there leaves the larger prize untouched.
Optimization is a different discipline. It asks not "can the machine do this on its own?" but "what is the best this process can do, and how do we hold it there?" That question is where I've spent most of my career — loop tuning, advanced process control, model predictive control on cement and polymer plants.
Automation answers can the machine do this. Optimization answers what is the best this can do — and how do we keep it there.
It's the same question for AI
Most organizations adopting AI today are automating: take a task a person does, hand it to a model, save the time. Useful. But the real value shows up when you stop asking the model to imitate the current process and start asking what the best version of that process looks like — and then design the system to hold it there.
That's a control problem, not a model problem. It needs setpoints, feedback, and a way to detect drift. It's the exact muscle you build optimizing a plant: you don't just make it run, you make it run well, and you keep it there when conditions change.
Written by Usman Nasir — control systems engineer, Stockholm.