This article originally appeared in KMWorld Magazine.
Pragmatic Steps to Deploying Next-Generation, Search-Based Applications Called “Intelligent Assistants”
I recently went to a prominent technical web site’s self-help application for my smart phone. It is a sad tale of low quality service and unnecessary costs that started with a simple search about leveraging fee-based service for my new phone.First, I queried using Google. I got a long list of postings on different sites that I had to click through. Two pages of links did not appear to have the answer I needed, so I went to the service provider’s site for help. I executed my query on the site and received, “no search results found.” So, desperate for an answer I called the support line.
Ironically, I was offered a recording of how I could solve the problem myself online. Frustrating, wouldn’t you say? After 10 minutes an agent answered. I repeated my problem. She told me she couldn’t solve the problem. I had to deal with application provider … and so the story goes.
This is happening every day inside of companies:
Prospects or consumers can’t find product information
Existing customers/consumers fail to find answers on self-help
Customer support and service technicians are forced to read through long lists of documents and disparate information sources. Sometimes they can’t find answers, so they rely on tribal knowledge.
R&D engineers invent their own answer to problems already solved… worse yet, they waste time on unsolvable problems or ideas that have already failed.
Historical practices underlie most of these challenges.
Document Focus – Historic methods for producing content have left companies with large stores of long documents. Content is often still optimized as though we’re using paper.
Not Purpose-Optimized –When the job of search solutions was viewed as a way to locate relevant documents, content structure was not optimized to allow search to locate specific content within the document. Islands of Information - Separate social media solutions are used for content such as tips, tricks and discussion forums. Some data remains in transactional systems. the result is separate sources of separately tagged information to manage, curate and search.
Lack of Insight on Use or Value - Metrics used to optimize consumer website experiences are not being used to optimize the productivity of the most expensive, critical knowledge worker tasks.
So what is the end result of these gaps in our knowledge management solutions? At a major electronics manufacturer, it meant that field service technicians were spending 15%of every week searching through disparate systems to find answers to repair challenges. That added up to $7.5 million to $10 million annually in excess costs in the system and critical down time for customers in their manufacturing process.
What is the answer? A new form of Siri-inspired search applications called, “Intelligent Assistants.”
Siri and Google have opened our eyes to the potential that if we can understand more about context and more about the details of the content, we can get beyond search sets that list documents and go straight to answers.
An Intelligent Assistant is a search-based application that sits on top of content that is established in smaller components, Component Content. Essentially, you must break down your knowledge into smaller chunks of information, typically by task, concept or idea Each smaller component can now be tagged, making comments about each step in a diagnosis or some smaller part of a product reference now available as a search result. These smaller pieces can now be easily scanned, consumed and used in real time.
For example, at a leading national channel-driven services organization, product line executives had to find a way to help their indirect channel to better understand and configure a complex range of higher end products for business customers. Training didn’t really help. The line of products included higher dollar products that were not sold every day. The product fit was subjective, and multiple options could be presented, making it too complicated for an auto-configurator.
It was far too easy for the sales reps to forget all the provisions and alternatives between sales calls. One of the greatest barriers to sales through a channel is perceived complexity. Channel organizations will avoid selling anything that is too complicated to learn. So, support costs escalated as the service organization “held the hands” of the sales force through a high-cost, deal-by-deal support model.
“Our valuable call center resources were squandered on answering basic “how-to” questions and FAQs. We had invested in world-class content that provided all the answers, but our search apps were straight out of the 1990s. The fastest route to an answer was to pick up the phone – even if that meant being on hold for 10 minutes!” --Senior Operations Executive, Large Global Manufacturer.
An Intelligent Assistant was created in only 4 months to guide configuration of the right product solution and to literally answer questions about product applicability and selection. Early results are very positive, and, with proper content curation, the organization believes it will reduce support costs by nearly 80% and be able to redirect call center resources to growing revenue.
How do You Get Started?
If you take careful steps up front, you can quickly ramp your Knowledge Management team to create Intelligent Assistant applications throughout your organization to solve your most pressing collaboration challenges.
Following three simple techniques to take your first steps toward Intelligent Assistants:
Pareto each Problem to drive effort and the business case
Establish a Task-Focused Component Content Architecture
Leverage DITA-based Technologies
Pareto each Problem
According to Wikipedia, “The Pareto principle states that, for many events, roughly 80% of the effects come from 20% of the causes. Business-management consultant Joseph M. Juran suggested the principle and named it after Italian economist Vilfredo Pareto, who observed in 1906 that 80% of the land in Italy was owned by 20% of the population; Pareto developed the principle by observing that 20% of the pea pods in his garden contained 80% of the peas.”
So, it is likely, looking at any given problem, that having proper component content over 20% of the content is going to lead to most of your end results. The trick is: which 20%?
Usage isn’t a bad way to start. Under this approach, you analyze which content is currently used the most, select the 20% highest use and create an Intelligent Assistant, with that 20% most carefully re-purposed into detailed components. That may help, but it can also miss the point. Current usage patterns can reflect engrained behaviors, and users tend to stop searching for content they can’t find. You could select content that is easy to find, but not select content that solves your problem or achieves your targeted benefit.
A better way to start is to analyze barriers to your business objective. We strongly suggest beginning your project with a clear, measurable objective in mind. Then, analyze at the task level where 80% of the potential time or cost is imbedded. Once you’ve found the fewest tasks that comprise 80% of the performance gap, take a look at information usage, research time, reading time, and searching multiple content stores.
As an example, more detailed benefits analysis leveraging Pareto techniques found the following detailed issues that Intelligent Assistants could resolve for a leading global electronics manufacturer:
50% of their time was spent on research.
Getting the research right and complete the first time meant only having to suit-up once to enter the cleanroom then solve the issue.
First-shift-solve was important. Getting the right parts identified and ordered before the next-morning shipping cut-off was key to minimizing customer downtime.
Making sure the information could be printed on clean-room paper as well as made available electronically was key. Not all customers would allow a smartphone / laptop / tablet into their high-security fabrication facility / cleanroom
Get the Information Architecture Right
Two primary barriers usually must be overcome to deliver Intelligent Assistant applications that hit targeted results:
Define smaller more precise units of content,
Integrate across multiple, fractured content stores.
So, having identified the content that is needed to solve your Pareto analysis, you must define a taxonomy and metadata that adequately identifies your more detailed components. The best way to do this is at the task and role level.
Defining your content in smaller chunks around roles, tasks and concepts unlocks your ability to move to Intelligent Assistants. When procedures, bulletins, and training material are in smaller chunks, I can more accurately tag them and make them searchable. These smaller chunks are also easier to summarize and present as answers, eliminating the need to read through long documents looking for the paragraphs that matter.
Very frequently, an Intelligent Assistant must also integrate information across separate data stores and content types. For example, the Intelligent Assistant that drove significant productivity improvements for a major electronics manufacturer had to integrate results from multiple sources of technical publications, ERP systems, parts systems, and several service procedure systems. In total, over 15 systems were integrated to present answers to technicians, versus returning sets of documents to read.
The benefits possible through Intelligent Assistants often unlock executive approval to get started on a refresh of your enterprise information architecture. The right IA allows you to leverage component content and integrate information across disparate sources of information to present answers. Answers lead to hard dollar benefits. Using Pareto analysis focuses the team on smaller bite size projects that deliver benefits. Benefits build momentum. You gain the right to refresh your Enterprise Information Architecture (EIA) one step at a time. If you don’t have an EIA, Intelligent Assistants are powerful ways to unlock executive commitment to create a high value EIA.
Leverage DITA Approaches and Technologies
The Darwin Information Typing Architecture (DITA) standard and related technologies are opening up a much more productive way to represent component content and create that content with simple DITA-based authoring and management tools.
Using DITA as the basis for component content has a number of advantages:
DITA is component oriented, so it is natural to create content components at the task or concept level.
The latest generation of DITA editors provide simple-to-use tools that business people can use to create DITA content, with no special training. You don’t have to be an XML expert to create DITA content.
Component-oriented review and approval speeds time-to-delivery for new content. You only review and approve new or revised components – not entire manuals.
DITA single-source content can be formatted for print or electronic delivery. This includes text-to-speech applications.
DITA uses semantic markup – a step in a procedure is actually tagged as a step versus a numbered paragraph. This improves search precision – for example, you can search for all the steps that refer to a specific error code.
As an XML format, DITA is machine-readable, so an application can interact with DITA content in intelligent ways. For example, DITA allows you to build an application that walks through procedures a step at a time to prevent users from getting lost in long, complex tasks.
Don’t Forget Governance… both Content and Value
Your new Intelligent Assistant application needs governance on an ongoing basis. The more detailed components must be maintained and refreshed. Often the first step unlocks ideas for Intelligent Assistants in new areas of the organization, so you’ll need careful governance over your evolving EIA to maintain the value of existing solutions. Governance over the content and metadata will be required to ensure ongoing usefulness.
As part of your business case, determine if you would benefit from automation for your content intelligence. A new line of products are helping companies simplify and automate governance and extension of robust knowledge applications like Intelligent Assistants. For example, Smartlogic, can help you leverage your new information architecture to extend and automatically tag new content, dramatically simplifying the maintenance and upkeep of your EIA.
Don’t forget governing the value of your solution. Build your Intelligent Assistants as web-enabled applications, and use the same web metrics your company uses for their corporate and eCommerce sites. Those same metrics tell you when people are searching too long, not finding what they need, or abandoning pages too early to have learned what they need. These metrics become strong hints of gaps in value or of the next part of the problem to solve.
Intelligent Assistants that deliver answers instead of long lists of large documents to be read can deliver significant bottom line value for knowledge workers and for your customers. While it may seem a daunting task, a few careful, pragmatic starting steps will help you deliver your first solution quickly.
Many times, you can use the same tools you already have. SharePoint and other leading collaboration tools have most of what you need. Along the way, considering DITA-based tools and some automation from governance and extension of the solution may be wise, if they can be cost-justified.
The benefits you deliver will unlock new momentum for knowledge applications in your enterprise, and Knowledge Management as a discipline can become a more strategic partner in achieving important business goals and objectives.
For a look at how we use information architecture to design and build an Intelligent Virtual Assistant download our white paper: Making Intelligent Virtual Assistants a Reality. In this paper we show how to deploy intelligent applications in your enterprise to achieve real business value.