Cognitive Computing - Part 2
Applying cognitive computing to KM

“Cognitive computing” is one of the biggest buzzwords in business today. Leaders are scrambling to understand what cognitive systems actually do, but few are clear on the impact those technologies will have and how they can be applied to address existing needs within their organizations.

In 2016, member-based nonprofit APQC assembled its 9th KM Advanced Working Group to explore the practical potential of cognitive computing in the knowledge management field. Over five months, KM leaders from APQC, Deloitte, EY, NASA, Pfizer and the U.S. Army Training and Doctrine Command evaluated both existing capabilities in the marketplace and what’s on the horizon. Ultimately, the group identified six areas where it believes cognitive technology has the most potential to transform KM systems and approaches.

The first article in this series explained the Advanced Working Group’s analysis of basic cognitive computing concepts and how it believes the technology will influence the future of knowledge work. This second installment takes a deeper look at the relationship between cognitive computing and knowledge management, focusing on the six use cases where cognitive systems show the most promise: content curation, search and discovery, expertise location, lessons learned analytics, data visualization and intelligent personal assistants.

Content curation

In many ways, the field of enterprise content management faces more daunting challenges than ever before. Along with the proliferation of traditional document types, employees are increasingly generating new forms of content, from wiki articles and social media conversation threads to graphics-rich presentations and YouTube-style videos. Content management strategies aim to curate enterprise content so that search can elevate the best materials in response to a user inquiry. But manual curation is costly, and most organizations’ efforts are only somewhat effective at getting the right content in front of users when and where they need it.

Machine learning can automate content curation by using algorithms to scan content items, find similarities between sources and cluster them in logical groups. The content then can be tagged accordingly and displayed both proactively and on demand in response to search queries. The goal is to improve the efficiency (and cost) of curation processes while eliminating the “misses” where relevant content exists, but the system fails to connect users with the best resources.

It is unlikely that machine learning will rival a human at curation, but it could certainly reduce the burden by automating tagging, clustering and delivery. The algorithms also have the potential to help organizations better understand their content and pinpoint areas where they have insufficiencies or gaps. Most manual curation relies on organization-centric taxonomies and ontologies (e.g., for business units, functions or product lines), whereas cognitive systems are better positioned to look outside those boundaries and identify similarities in content from diverse sources and contexts.

In addition, cognitive systems can support dynamic content structures that adapt to and reflect ongoing changes in vocabularies and knowledge domains. That may provide benefits in terms of breaking down knowledge silos and overcoming the inherent shortcomings of traditional fixed taxonomies.

Another advantage of automated curation is that KM teams would have less need to beg the business for experts’ time to support content classification and tagging projects (at least when using unsupervised algorithms). If experts’ time instead could be reallocated to more value-added activities associated with tacit knowledge capture and transfer, then the follow-on impact of such systems could be profound.

Search and discovery

Search is the web’s killer app, thanks to Google, Bing and other commercial search engines. Applying sophisticated algorithms and machine learning to billions of search queries per day allows those applications to set the bar very high for enterprise search. Yet, despite great strides, employees still complain that they can’t find what they need.

In some ways, enterprise search has a tougher challenge. With a much smaller volume of user activity data to draw on, an organization must understand what content it has and then tag, filter, contextualize and personalize that content to facilitate a particular employee and the flow of work.

The biggest complaint facing enterprise search today is filter failure: the inability to give employees just the information they are looking for, just in time and in the context of their specific roles and work processes. For most organizations, surfacing content is not the problem; instead, searches return too many results, making it difficult for users to identify the most critical or pertinent pieces of information.

The possibilities for cognitive computing to enhance enterprise search and discovery processes are enormous. While many intriguing opportunities exist for cognitive applications, 40 percent of KM professionals believe that search carries the most potential in the context of KM (see chart on page 19.KMWorld, February 2017, Vol. 26, Issue 2 or download chart).

Current search applications can traverse unstructured data and perform some automatic tagging, but the capabilities are limited. As search functionality evolves to incorporate cognitive capabilities, it will be able to provide more comprehensive answers—even when that involves interpreting images or videos or combining information hidden in multiple documents.

The enterprise applications will be powerful and have a potentially huge return on investment. For example, users will be able to ask the system questions such as:

• “When is the first available time for Joe, Mary and Sam to meet with me for one hour?”
• “When was the budget report last updated?”
• “List the parts in the electrical system.”
• “What are the safety requirements for storing acid?”
• “How many parts failed inspection in yesterday’s run?”
• “Do you see any patterns in last month’s lessons learned reports?”