Learn it from mistakes
Our nervous system needs to be trained. Most neurons do not yet know whether to send the electric shock left or right (and we only use about 2% of our entire brain). When we learn to walk, we'll fall over and over again, until our brain has found or created a neuron path that allows us to stand. At some point in time we even know what not to do to stay standing up, but in most cases this is done autonomously without us even knowing it. We'll probably never get insights in the actual neural path that leads to a decision with a certain outcome and even if we would be able to create a decision diagram, we wouldn't be able to comprehend the number of variables herein. Our brains are like a black box, very efficient, fast and enormously powerful.
Can we use this principle to our advantage in computer systems? Yes we can, even more so, some is already applied in most information systems through the use of a CPU or processor (equivalent of a human brain). Other people that write software determine for us what the capabilities are and what is 'learned' by this brain but we can also train it ourselves in what we call a neural network.
Most of us have somewhere in our careers performed questionnaires and used analytical software to perform analysis to find variables with a certain degree of certainty. We analyze a set of multiple choice questions and the respondents answers to determine if there is a relation between A and B, e.g. Do you like the environment and how likely is it that this fact drives your product choice? The result of the analysis might be: there is strong positive relation between our attitude towards the environment and our choice of products that are environmentally friendly.
In this case we know the variables but what if we don't, like most autonomous decisions in a human brain? Through a neural network in a computer we can put in large volumes of transactions, with any given number of variables (relevant or not) and the system will determine the outcome. How? Well, in the same manner as our brains: Through training ;)
Consider the following example (now it becomes relevant for BWise :)). I travel a lot for my work and fly around the globe. I use my Amex credit card for pretty much anything. Food, drinks, flights, atm's, hotels, taxi's etc. One of the most important risk of any credit card company is the risk of fraud, since most of that is covered by them. They need to make sure they do everything to prevent fraud because bottom line, that will cost them money. There are many variables in my payments using my credit card that could indicate fraud, but Amex has to pick those transactions that are most certain to be fraud (the sheer number of transactions will otherwise make it unmanageable). Variables include type of payment, receiving party (hotel, cab or pizza delivery guy), amount, local time, location in the world, total outstanding payments, time between payments, similarities to other payments etc. These variables show different behavior from person to person since they live different lives. So it is difficult or close to impossible to create a model that produces a 99.999% reliable overview of fraudulent transactions.
Based on historic payments and fraud cases found, Amex could create a model that can indicate possible fraud. With a computer and a database, you can use all historic data to produce (train) a brain (neural network). After training, any given transaction could be put into the brain and it would produce the most likely result (either fraud or no fraud) with a certain confidence.
Think about the possibilities of this in the GRC space! There are many but they will all require historic data sets to 'train' our brain. We should start building those now.... So that WHEN BWise is going to use it in the future we'll all be ready. Also consider the options of a collective brain (Amex, MasterCard and others all-share each other's historic fraud data to create an even more accurate brain from which all will benefit). The future is now.... :)