Years of ground-level perspective mean I know just how important it is to rapidly build internal trust and create fast value through pilot projects.
Quick wins are crucial. But they’re not the whole picture. Being truly holistic means taking a dual approach, having a long-term transformation's "North Star" AND changing your actions today (that's how mindset shifts happen, remember?)
What we often hear: "I do not want to do more assessments. Let's stop doing it on paper, I want to go to real activities"
It’s also crucial to be clear on what constitutes a ‘win’ and understand how difficult it is to make predictions or truly reshape existing processes. I can think of a recent example using AI technology that demonstrates this.
A group of researchers from ABB and Boliden used AI simulations to optimize concentration processes. What constituted a successful result – a win – was well defined in simulation environment. The goal of the project was to check if AI method called reinforcement learning could perform better then traditional regulatory control.
The algorithm successfully taught itself to accomplish the task, but it turns out that it did so by heavily exploiting a small model mismatch between the simulation and reality. Hence, the obtained control strategy wouldn’t work in the real world. So the next step was for the team to change the goal and optimize the overall profit instead of optimizing the tracking of a process variable.
With the new goal in place, the AI algorithm took 80 hours of simulation to learn to run the process profitably. That was the equivalent to a plant operating time of more than 300 years.
Sounds impressive, but further studies are still needed before we get to a viable production solution.
This example shows the difficulty of making predictions for a real production environment even with the most powerful analytics tools, and it stresses the importance of having the skills to understand, question, and rethink existing processes.