Regular readers of my blog (as well as those who hear me speak and our Decision Management clients) know that I talk alot about automating decisions. Really, a lot. The evidence that taking control of decisions, specifically operational decisions, and developing Decision Management Systems to automate and manage those decisions drives improved organizational effectiveness is extremely compelling.
Regular readers of my blog (as well as those who hear me speak and our Decision Management clients) know that I talk alot about automating decisions. Really, a lot. The evidence that taking control of decisions, specifically operational decisions, and developing Decision Management Systems to automate and manage those decisions drives improved organizational effectiveness is extremely compelling. The stories in my book, the stories I discuss in my presentations, and the stories you hear from vendors in this space are of increased agility, decreased costs, increased profit and decreased fraud. Decision Management and Decision Management Systems work.
Despite these proof points I still get some push back. For too long the association many business people have is that something automated will be inflexible, costly, hard to adopt and one-size-doesn’t-really-fit-anyone. They worry that completely automating decisions will require compromises in accuracy and devalues their staff. They associate automated decision with some kind of big black box that issues decisions that must be obeyed – something like Hal from 2001 but with less humanity. Which brings us to the misconceptions
- Decision Management Systems are just like other systems
Decision Management Systems are not systems of record and are not responsible for the workflow, the process, of your business. They just make decisions. This alone would make them different from the systems that store and manage the data that your business relies on but they are also built differently:
- They are agile, built using a business rules foundation to make sure that they way they work is transparent and can easily be changed when necessary, often by non-technical people on the business side of the house who understand when such changes are required.
- They are analytic, using analytics not as a way to report progress or analyze results but as a driver of accurate, precisely targeted behavior.
- They are adaptive, learning from what works and supporting their users as they experiment and learn themselves.
- Automating a decision means automating it 100% of the time
Some Decision Management Systems do automate a decision 100% of the time, handling the decision every time it is required. This is particularly true when the channel that requires decisions is completely automated such as a web channel or a kiosk. Clearly in those circumstances only 100% automation makes sense. But in many other scenarios automating a decision means handling 80% or 90% (or even 95%) of transactions while referring the rest to a human decision maker (with some explanation of why each is being referred). This latter kind of system is much more common and has the added advantage that you can start simple, handling perhaps the easiest 50% of the transactions, and gradually adapt the system to handle a higher percentage over time. Business rules management systems enable this kind of iterative development and its very effective in Decision Management because the overall system already knows how to handle the manual decisions.
- Automating a decision means automating 100% of the decision
Similarly some Decision Management Systems do make 100% of the decision – the process reaches the point where a decision is required, the Decision Management System makes the whole decision, and the process continues. In many other situations, though, the system makes only part of the decision. one of the reasons I really like Decision Modeling is that it allows you to take a high level decision (where automation might be impractical or heavily resisted) and break it down in its component pieces. Some of these will lend themselves to automation with business rules, others might be clearly handled using an analytic model. Others may require human judgment. With such a model in hand it is possible to determine the boundaries of your automated decision. It even allows you to define how a largely automated decision might require human judgment at times.
I would also point out that many automated decisions are neither 100% of decisions NOR 100% of each decision. Such systems are not pure Decision Management Systems nor are they pure Decision Support Systems – they are a bit of a blend. Nevertheless they can be very effective.