@ Supply Chain Management

This is an interview with Pierre Haren, CEO of ILOG prior to its acquisition by IBM on what decision management is and how it came to be used in various situations. A look back (and forward) at 20 years of Decision Management.

This is also my particular field of expertise and Pierre alludes to the case of the product that I work with gaining acceptance in IBM’s wafer fabrication facility in East Fishkill, NY.

We had a great moment when we were able to successfully deploy a complex mix of ILOG and IBM tools at IBM Fishkill, in what was considered the most modern and fastest changing semi-conductor plant in the world. At that time, before our acquisition by IBM, IBM was a reluctant customer. I visited the head of the plant and promised that we would install the complete system for free. He would only pay if he kept the system in operation for more than a month.

Understandably, he was skeptical that a small decision management company could improve on the work of hundreds of excellent engineers who continuously tuned the $4B plant. We deployed a system which every five minutes re-computes the optimal plan for the whole plant for the next eight hours. It combines real-time data from the MES system to rules-based processes, an optimization-based scheduler, and advanced graphics to enable the plant operators to "see the current future" for the next eight hours.

The first day’s objective was low: do no harm. Within a month, all the operators where convinced and we were paid. More important, we changed forever the balance of what computers do and what operators do in that plant.

What’s important about this confluence of components i.e. Rules Management, Optimization using CPLEX and Visualization as it used in that fab (and since then in other fabs in the world) is how such a system can be used for execution. Not planning, Not strategy. But execution on a near real time basis – that means that actual tools on the fab floor act on the recommendations produced by the decision management system which in turn sees the action of the tools and then uses that as an input in further decisions. When operators and managers assert manual control in the operation of the tools or fab in general, the system accepts that and continues without issue. So the system takes into account the dynamic nature of the fab, recomputes and proceeds. I’ve blogged about this system : FabPowerOps in earlier blog postings – take a look.

Now I’m hoping to take this to the next level, deploying a federated system of such applications that not only interact with the fab floor as well as human operators, engineers and managers but with other apps taking their inputs and working on that.

Great stuff, I assure you!!