“Bring in the strategy robot! Move over and give him a seat!
What does that thing do… you ask? He’s here to represent the employees. Around this table, we only have room for 15 executives to discuss the strategy, but we value our 100,000+ workers on the front lines and our millions of customers. We want to bring their voices directly into the strategy conversation.
The robot has been scanning the millions of social media comments, conference call recordings and documents from our intranet… looking for insights. He knows all about our strategy and our performance, of course.
Is he accurate, you ask? Not 100%. He’s comparable to the best social media sentiment analysis systems, at 80-90% accuracy. Most of what the robot says is relevant. We’ll give him feedback, and he’ll get ‘smarter’ over time.”
OK – maybe West World is not ready to host your strategy meeting yet. In data science terms, we are simply talking about a machine learning “classifier” that is trained to determine if content or posts are related to our strategy. With gigabytes of employee knowledge being recorded on collaboration platforms such as Slack and Skype, we desperately need the robot to separate out the “noise” and find the insights. Otherwise, all of this “working out loud” data will not become corporate-wide knowledge.
The robot-classifier could organize the output around the logic of the corporate strategy map (see below). The goal here would be to match customer and employee-generated insights with strategic objectives and targets. For strategy managers, this can bring greater understanding of what is going wrong (or right) with the strategy and the execution. For example, why didn’t hiring more people improve operations?
Note: The strategy map on the left is based on Kaplan-Norton’s Balanced Scorecard & Strategy Management System.
The scenario described above uses “supervised” machine learning, where the classes (the buckets for classifying the content) are defined in advance. The robot-classifier simply makes the best decisions when scanning the content (for example, finding data and knowledge nuggets that fit the class = “inspire loyalty”).
In addition to the supervised classification, we can run “unsupervised” clustering on the same employee and customer data. Instead of telling the robot the classes, we can let the robot make its own clusters, and build a strategy map based simply on what employees and customers are saying. This would result in alternative, “bottom-up” strategy maps. Generated by employees customer and yes… the robot..
What’s the value of an unsupervised, robot-generated strategy map? For most organizations that use the “top down” traditional strategy maps, an auto generated bottom up strategy map would be an interesting counterpoint to generate discussions. Are these in fact good ideas that are bubbling up from the employees and customers? Organizations that want to become more agile and spread decision-making towards the front lines should embrace these bottom-up strategy maps.
The robot – configured to classify or cluster the strategy - would initially be just (selectively) repeating ideas expressed by employees and customers. This filtering and “bubbling up” of insights is critical. Eventually, the robot might truly generate ideas of its own. The ideas might even be good ones!