Cognitive Computing - Part 2
Applying cognitive computing to KM
The ability to pose such queries and receive explicit responses in narrative form will not only save employees time searching for information, but will also unlock access to information and insights that are currently unobtainable (either because the patterns are undetectable via human analysis or because finding the answers manually would be prohibitively complicated and time-consuming). And once organizations start leveraging cognitive computing at the optimum level, users won’t even need search for routine tasks. Predictive analytics and machine learning will take over and anticipate what they need. The system will analyze all the contextual clues available to it—the user’s skillset and role, his or her current projects and location, the communities or networks the user is active in and everything the user has searched for or looked at in the past—and combine that with domain knowledge to recommend information and updates it believes will be relevant.
Many of those capabilities are available currently, but they are pricey and out of reach for most organizations. The Advanced Working Group anticipates that as the tools become ubiquitous and vendors offer more cloud-based options, prices will fall and adoption will rise. KM professionals should start thinking through the possibilities so that they are prepared with a strategy when the capabilities come within reach.
Expertise location
Employees are constantly seeking experts and expertise to help them solve problems or make them smarter about a topic. What is the best way to search across a large, complex organization and find people with the right knowledge? There are many options, including reviewing internal directories or SharePoint lists, contacting the authors listed on key documents or asking for recommendations in related communities or social networks. But for the past decade, enterprise efforts have focused on “expertise location systems,” most of which rely on employees creating LinkedIn-style profiles that list their areas of knowledge, interests and past and current assignments.
Expertise location systems can provide a lot of value, but there are limitations. For one, the data typically is compiled from a variety of sources (e.g., talent management, HR, project scheduling and assignments), each of which provides only a small sliver of the overall story. That makes it difficult to integrate enough inputs to get an accurate, holistic picture of someone’s expertise.
Furthermore, most expertise systems require employees to fill out and then manually update information about themselves. It is difficult to convince employees to fill out profiles, and even when they do, the shelf life of such data is limited. This means that profiles often do not contain the most current—and therefore relevant—information about a person’s experience or interests. In short, expertise location strategies and systems have been stalled for over a decade waiting for the underlying technology to catch up—and it finally has.
Cognitive systems have the potential to scan numerous disparate sources of information (everything from official role information and project assignments to content contributions, downloads, emails and social networking threads) to identify subject matter experts or people who work on specific topics. Such systems can improve the quality of colleague recommendations, reduce the time required to find help and answers, and uncover hidden pockets of expertise—all while relieving individual users from the burden of profile upkeep.
The model for next-generation expertise location may resemble online dating, if it were powered by Netflix or Amazon’s “If you liked X, then you might like Y” recommendation engine. The vision is for a system that looks at all the available information about a particular user (including his or her actions in the organization’s digital environments) and then proactively recommends colleagues with interests in common.
More advanced cognitive systems might even learn employees’ preferences and emphasize the specific expertise markers (e.g., official resume experience, level of community involvement) most important to each seeker.
The goal of those super-charged cognitive expertise locators is to create a better networked organization in which colleagues can find someone with relevant expertise, knowledge or interests, even if they work in different business units or on different continents.
In addition to the collaborative synergies offered by such boundary-spanning relationships, there may be logistical advantages for organizations with a limited number of “known experts” on key topics. By broadening the available pool of knowledgeable contacts and connecting people to an array of contacts with expertise for a given need, cognitive systems may relieve overburdened experts from having to answer every question about a particular topic—especially novice ones that can easily be answered by mid-career professionals who work in the field. In theory, this will free up expert time for strategic thinking, innovation and tackling the subset of questions and problems that truly require their attention.