Game-Changing AI Breakthroughs in KM

Yesterday I participated in a round table for KM World called “Game-Changing Breakthroughs in Knowledge Management”. I joined Heather Richards, VP KM Product Strategy at Verint and Vivek Sriram, Chief Marketing Officer at Lucidworks to discuss how discovering the value in corporate knowledge is amplified by new technologies, such as AI, text analytics, cognitive search, machine learning, natural language processing, mobile email management, and new policy efforts.

Downloads

You can download the whole program at http://www.kmworld.com/Webinars/1258-Game-Changing-Breakthroughs-in-Knowledge-Management.htm or check out my slides:

The promise of AI

Cognitive computing, machine learning and AI are all used interchangeably to describe a suite of functionality that holds huge promise for changing the way we work. In fact, 80% of CEOs believe AI will change they way they do business in the next 5 year.  But they are finding that it takes longer and is more complex than anticipated.  Unfortunately, and contrary to what one senior manager hoped, you can’t just buy AI off the shelf, install it and sit back and watch it work its magic.

In fact, a successful AI implementation involve many domains and disciplines.  Not only do you need to understand machine learning technologies like natural language processing and semantic graph analysis, you also need a huge computer processing infrastructure, large amounts of data to analyze and specific use case expertise.

Considerations for implementing AI

The first step is to pick a problem that you can apply AI to.  There are three main things to consider when sifting through potential problems to apply AI to solve:

The first is – can it be solved with the techniques and resources available today?  Second, and more important, is the problem central to business? Thirdly, can the problem be solved any other way? Once you have identified an problem that is important and solvable using AI, then you need to coordinate the various domains and disciplines required for an AI implementation as well as the non-machine learning components that may be needed to distribute the application you develop.

Northern Light has the expertise in machine learning technologies and a huge computer processing infrastructure and access to large amounts of data to analyze.   I presented a few use cases we’ve created with our customers.

Getting insights instead of searching

Here’s the first problem – people don’t like to search and read through results lists, particularly millennials. The average millennial switches social media platforms 27 times an hour.  They hate searching and they don’t read long lists.  They just want to get an answer. 

Imagine you want to learn more about the future of fuel cell powered cars.  In a regular search experience, you might type fuel cell cars into a search engine and see what you get – which would be a list of hundreds or thousands are web site or reports that mention fuel cells and cars in date or relevancy ranked order.  But, what if you could get a report that tells you what all those documents have to say about fuel cell powered cars?  With Northern Light’s SinglePoint, you can get just that with our Insight Report.

Here we use AI to read every document in the search results and provide a summary of the most important sentence in the documents to give you an readable overview of the whole result set. 

This allows occasional researchers to learn more by reading an Insights Report than they would be able to by engaging in the traditional search process of scanning search results and picking one or two documents to download and read.   Power users can save time by reading the Insights Report to learn what major issues are uncovered by their search and then drill down into key topics of interest.   They can also use Insight Reports to publish key information to the organization in newsletters and on dashboards.  This is a point of radical discontinuity in the search process – it changes everything.

AI for Social media marketing

How can you tell if hashtags or keyword are used by the same consumers in a target market?

For example, for example which hashtags (and keywords) co-occur with cancer? You could search, say twitter, for both and do a manual inspection, but 51000 hashtags co-occur with cancer and more are added everyday. 

There are tools for determining co-occurrence, but the results can be less than useful. With our Social Analytics tools, we use Machine learning-based semantic analysis to determine which hashtags and keywords are actually related to each other in a body of social media posts.  To do this, we build a training model of known good tweets that are about the topic of interest.  Then we use machine learning to semantically analyze the tweets in candidate hashtags to find those whose content is similar. Then we use the similarity to help inform market research and social media marketing activities. 

How does an organization capture critical intelligence learned in the field, at sales calls or conferences for example, and a act on that intelligence? Personnel attend conferences, go on sales calls, have conversations and learn intelligence about competitors, and products, and technologies.  This intelligence is (sporadically) emailed around or talked about, but not curated, searchable, shareable, or actionable. 

Gathering field intelligence

Our question was, can chatbot technology capture this intelligence and make it easy to bring into the knowledge management process? The application we build is an app that an employee can use while out in the field.  They answer a brief survey, review the transcript, add any additional supporting files, like a picture of a poster or brochure, and then submit the information directly to the research portal for curation and potential action. Once it submitted, the intelligence can be routed through a purpose built workflow.