Is Machine Learning the Next Major Step in ECM?
Enterprise Content Management (ECM) is defined by the Association for Information and Image Management (AIIM) as “the strategies, methods and tools used to capture, manage, store, preserve, and deliver content and documents related to organisational processes”. Implicit in this definition is the intersection of people, processes and technology; each of those three elements must play its part for ECM to operate.
However, it’s likely that in the very near future another essential element will come into play: intelligent machines.
The Rise of the Machines
While it could be argued that this would fall under the “technology” aspect of the ECM trinity, our view is that it would in fact be more akin to a separate, fourth element. The introduction of true machine learning into the equation would bring new capabilities and benefits to ECM that would potentially be closer to what is currently delivered by people, rather than technological resources.
Machine learning allows computers to learn by carrying out tasks, rather than simply following explicit programming. One simple example of machine learning that most of us will have had some experience of is spam filtering. The filtering software learns, via input examples, which data features – certain words or combinations of words, for example – are more likely to be spam. It’s a straightforward but (usually) effective method of classifying incoming emails or form data – in this case each piece of content is either “spam” or “not spam” and can be sorted accordingly.
Benefits of Machine Intelligence
It doesn’t take a great leap of imagination to see how this type of machine intelligence can be of tremendous benefit in ECM applications – software capable of intelligently classifying, tagging and filtering data, whether that takes the form of business files and archives; incoming customer correspondence; employee or customer records; or financial or audit/regulatory compliance data.
Machine intelligence has particular potential for tackling two of the biggest scourges of current data management: dark data (the mass of data that is accumulated and stored, but for whatever reason is not analysed or used to inform business decisions) and ROT (redundant, obsolete and trivial data). A 2015 report by Veritas estimated around 54% of all stored data is dark, and 32% ROT; leaving only 14% as business-critical.
Lack of adequate people, processes or technology is often cited as the reason for the accumulation of dark data and ROT; most businesses simply lack the resources to adequately manage the entirety of their stored data. While systems, workflows and even organisational culture may need to evolve to take advantage of this new element, a robust and agile machine learning solution could be exactly the additional resource required to allow businesses to fully exploit the potential of the data available to them.