Lew Platt, HP. “If only HP knew what HP knows, we’d be three times more productive.”
Much is discussed these days about Big Data and the role of Predictive Analytics to help create better business decisions and provide information from data. How would it be, however if we could apply Predictive Analytics to knowledge?
What I am getting at here is the role that Predictive Analytics can play in recommending information to you, suggesting next steps for you in a business process, connecting information together based on links elsewhere and generally providing suggestions based on modelled interactions which have been validated elsewhere in the system.
A knowledge repository, such as a file system, a document management system, a social business solution or just a good old discussion database is essentially a data set of unstructured information which, if modelled correctly, could provide inferences which could be applied to assist the user going forward.
Such an idea, of course is not new. I have worked on so-called Expert Systems in the past and most recently, of course, IBM has turned Watson to the cause of healthcare and cancer research.
IBM Watson is drawing inferences between huge data sets of data and information to help doctors with the work they’re doing in cancer research. How can we, however, use similar concepts (at a fraction of the cost) to help us with our rather less noble quests in every day business?
Simplistically we could implement predictive analytics by automatically linking text in a wiki with definition pages. e.g. If I wrote The quick brown fox jumps over the lazy dog and I have wiki definition pages for fox and dog, it would be great if the system would create links to these pages for me. Better still, the system could save me the bother of having to go to my own definition of fox or dog and could go out to the internet, pull the information in that I need and create that page for me. Of course, like Watson, I need to vet the information it comes back with – only I can truly know if the results the system is giving me are either appropriate or accurate. Using scoring, weighting and analysis, however we could teach these systems to get better over time.
If you are a user of the popular Zite application for the iPad you might be familiar with such a feedback loop already. You tell it what you’re interested in, it gives you news articles and when you strongly like or dislike an article you can tell it – it then adjusts the quantity of similar news depending on your choices.
Now suppose all this good work could be turned on the silos of information residing in your organisation. Suppose you employed predictive analytics, knowledge management and social business to built a learning, predicting and truly helpful solution for your business.
Suppose you’re writing an article about a new HR policy and at the side of the page the system recommended a blog post that a previously unknown colleague had written? Suppose it unearthed an email you weren’t previously familiar with or found a document on a file server which might help you. Consider too, such a systems ability to reduce re-work when preparing a presentation. Wouldn’t it be great if the hours you spent on that graphic for that slide only to discover that someone else has already created one – a better one – after your presentation – could be eliminated? We can do all these things just now through tagging, searching and the likes, but all these require our input at the time when the information is put in and also at the time when we’re looking for something.
Predictive Knowledge is something which really can only start off as Poorly Educated Guesses, but with time and the right feedback mechanisms we might approach a smart, learning solution which can infer connections between what we’re doing and what already exists – internally or externally – to help us work smarter.
As Andrew McAfee pointed out Lew Platt of HP recognised that
“If only HP knew what HP knows, we’d be three times more productive.”
How much more productive would your organisation be with predictive knowledge?
