Russian author Tolstoy wrote that every happy family looks alike, yet every unhappy family is unhappy in its own way. Translated into a business continuity context, the message is that enterprises and organisations that experience no disruptions to business have often identified the same core set of things to do properly – and which they do indeed do properly. Those whose operations falter or grind to a halt have neglected an essential aspect. However, the nature of that aspect may differ from one enterprise to another, or even within the same enterprise (for successive disruptions). There may be many points to be checked to avoid business discontinuity. Could machine learning speed the process and help to spot business continuity failures before they happen?
Machine learning essentially does what human beings do when they look for patterns in data. It then automates and accelerates the process. However, it can’t be smarter than the smartest of people, and not even the smartest have come up with a foolproof way of spotting all business continuity failures ahead of time. That said, machine learning has demonstrated some interesting capabilities in organisations, such as predicting when employees will leave, or when they might start to damage or steal from their employer.
While a machine can be taught to look for similarities in datasets, it can be more difficult to have it correctly pick out anomalies. In the example above, a disengaged employee can be characterised by certain actions or behaviour. A machine can look for this pattern and find a match. In business continuity, there may be many things that can cause disruption. At the moment, the approach is more about pre-identifying risky situations and having the machine look for any similarities, instead of trying to decide if something is an anomaly. In short, there is no “set it and forget it” solution for the foreseeable future. However, machine learning could be a useful tool to reduce risks to business continuity, even if it cannot eliminate them.