Technology Innovations - Artificial Intelligence

December 12, 2016

Avoiding Painful Customer Support Calls

By Edward Smith, Sr. Business Development Consultant, Canon Information and Imaging Solutions

Artificial intelligence (AI) and machine learning (ML) are among the most advanced technologies currently emerging from tech centers like Silicon Valley. As a technology scout for Canon, I help our R&D teams introduce such emerging technologies to Canon’s diverse product groups to maintain our leadership across a variety of product areas. An effective way to introduce these new technologies and stimulate use case discussions among solutions experts is to demonstrate how they might be used to solve a customer problem, or to enhance customer experience. A recent example has shown how AI and ML might be applied to improve the customer support experience, leveraging the foundational support of another technology: natural language processing.

Painful Customer Support Calls

When we can’t get a product to do what it’s supposed to do - it’s painful. Not just because it’s not working the way it should – but because we might have to spend a painful amount of time on the telephone with a customer support organization.

Most of us avoid calling customer support because of our multiple past experiences of seemingly endless Q&A sessions with representatives that have to run through long scripts of mostly irrelevant questions (Is the product plugged in? Did you turn it on? etc.). The point of all these questions, of course, is to help the support agent (who really does want to help) narrow all the possible solutions to the single most relevant documented solution to fix your problem.

To avoid this unpleasant and sometimes ineffective interrogation, many of us first seek help on the manufacturer’s support website (FAQs, forums, etc.). Unfortunately, entering a few keywords that describe our problem typically results in an avalanche of irrelevant results. In desperation we may even go beyond the first page of search results to try and find a solution - sometimes as far as the third or fourth page before ultimately giving up.

So – How come it’s so easy to find the exact facts or product information we’re looking for among all the websites and data on the entire world wide web with a search engine like Goggle, Bing, or Yahoo – but we can’t seem to find a single relevant document to answer to our problem on a manufacturer’s website?

The difference is that most webpages and web-accessible content have been carefully optimized with descriptive information, often manually, to ensure search engines can both find it, and understand what it is about (typically called search engine optimization, SEO). But most enterprise documents do not have any additional descriptive data (metadata) to describe what they are about in order to help search engines narrow the possible options down to the most relevant information in response to a search query. (When was the last time you added descriptive metadata to a WORD document?). This means the “old fashioned” keyword style search techniques often have to be used, and since most documents within a company have the same keywords – an avalanche of (mostly irrelevant) documents are returned, just because they happen to contain one of the keywords from your search query somewhere in the text.

Emerging Technologies Can Help

Natural Language Processing (NLP) is a technology that can help solve this problem. This technology has been steadily improving over the past several years as machine learning has automated its ability to interpret and understand the words and concepts within documents over the traditional, manually-administered rules-based processing techniques. At CIIS, we have implemented a technology platform that leverages the latest NLP technologies to augment document content with a rich set of metadata to clearly describe and categorize a document, and the concepts it contains.

With a knowledge base of solution documents that are fully described with this type of metadata, a smart search engine is able to filter candidate solution results much more effectively to help customers quickly find a documented solution to their problem.

This kind of fully augmented solution database can also fuel a new generation of intelligent “chat bots” that have started to emerge. One of these smart bots might pop up after a few unsuccessful solution searches within a self-serve support database. The bot would interactively ask highly relevant questions to appropriately enhance what a customer has been entering in the search box to help locate a solution – or possibly to prepare a problem summary that can be passed to a support agent if a solution can’t ultimately be found in the public support knowledgebase. Such a problem summary would eliminate the bulk of the Q&A session from a support agent, who could start with an exact problem description instead of asking for a re-explanation – and possibly have several highly relevant non-public solution documents pre-fetched for him based on the problem parameters initially discovered by the smart chat bot.

Everyone would benefit. The customer would either obtain the needed solution from the online knowledgebase, or would quickly receive the support he needs from a live agent who could join the solution search with an understanding of the problem already captured. This would, of course, also shorten the duration of support calls – making manufacturers happier.

This solution use case example has generated multiple conversations within different Canon groups for potential applications of the combination of natural language processing, machine learning and AI chat bots. As an advanced solutions organization, CIIS is always on the lookout for new ways to employ innovative technologies to benefit Canon’s diverse customers.