There’s a lot of of excitement about data, Artificial Intelligence, Machine Learning and generally analytics. Much of it is focused on the impact on the workforce and the machines taking over the world in some dystopian future. Another source of the excitement is from the vendor community who are rapidly commercializing propositions and products to placate the nervous customer community who are being told, and appear to be listening, to the proposition that they are being left behind.
My observations have left me to conclude that these propositions are largely technical in nature. Little is talked about the suitability of current business structures to reap the apparent benefits offered by these ‘new’ inventions. I purposely put ‘new’ in quotation marks because many of the techniques being touted as revolutionary have been around for decades. Geoff Hinton the ‘Godfather’ of deep learning has been working in this space for over 40 years. His progeny of PhD students are now the engines behind some of the most exciting AI based companies. What is new is that the availability of sufficient compute and storage at a sufficiently low cost point has turned many of the inventions into the innovations of today.
Returning to the point of this piece, I believe many organisations are inherently unstable and potentially flushing money down the toilet unless they understand the structural and cultural change necessary to benefit from these innovations. Andrew Ng, one of the credible voices of the new wave of AI evangelists made this point only recently. He pointed out that the successful beneficiaries of these innovations understand the complete system that is necessary to be a data driven, learning machine !
What is also obvious is that many of the theories behind system control are ignored by traditional companies and as such find that their implementations do not produce the expected benefits. I have been mulling over this for quite some time and as a result this piece is the first of a series which I hope will engage and generate some debate. To begin I want to share an analogue that will hopefully exemplify my point.
Imagine you had a personal target of driving down to the local supermarket before it closed in 15 minutes time. You know the route you have been there several times before. You get into your car and decide to conduct an experiment, as you want to employ the same methodology as you do at work when you gather data about your customers to build a model that predicts how they will respond to your email campaign. You start the car and set off, sampling your environment by opening and closing your eyes at 10 second intervals……… The rest would probably be history or at the very least a few points on your license and loss of no claims bonus.
The point here is that keeping your eyes open allows you to scan your environment, detect unforeseen events, roadworks, obviously other vehicles etc. However the sampling rate in traditional organisations is painfully slow and analagous to the opening and closing your eyes daily or even monthly. This sampling rate is caused by business structure, bureaucracy, decision making etc etc. Sampling theory was first explained scientifically by Harry Nyquist and Claude Shannon https://en.m.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem who proved that to accurately replicate an analogue signal digitally you had to sample it at a rate of twice the frequency of the signal. If you don’t you lose valuable information. Now, organisations are complicated with many environmental signals that affect them, however the base theorem still applies.
The nirvana is a continuous process. This is exactly how the autopilot in an aircraft works, taking into account the changing weather conditions, flight plans etc etc and justing the aircraft’s control systems accordingly. In the business world Amazon and Google are busily instrumenting the customer base with tablets, Alexas, and Homes which sample continuously and adjust their strategies accordingly. They are becoming giant learning machines. The implications are massive for their internal structure and business process. Amazon continually experiments, which turns traditional business planning and business cases on its head.
Applying the Niquist- Shannon principle to traditional businesses would typically result in a prediction of difficult to control, inherently unstable systems. A false sense of ‘stability’ in many businesses is then gained by damping the system using governance and a barrage of meetings to delay having to change course. Damping is a means of controlling a system by reducing its sensitivity to environmental signals. However, also in this case, the inherent lack of responsiveness can equally lead to instability.
So I would advocate that unless businesses recognize the need to build inherently stable systems they will rapidly go out of control and fail and the rate of change in the environment can no longer be dealt with using existing approaches.
What are these new approaches…..