All Posts

    Artificial intelligence (AI) isn't a new magic concept. 

    AI has been kicking around for a while now, though many associate the concept with Alan Turing, who introduced what we know as the "Turing Test" in a 1950 paper, "Computing Machinery and Intelligence."   

    What is new is the ability to scale human to machine conversations — once the realm only of humans — where the machine provides relevant answers to specific questions at an individual context. 

    We see this scale in the mass market thanks to entrants from large consumer electronics brands such as Apple’s Siri and Amazon’s Alexa, but this capability has yet to reach the corporate market. 

    What Will Push AI Into the Workplace

    AI-enabled search promises to transform the way people interact with information and digital assets, driving new efficiencies and creating value from information that has been all but lost in the “digital junk drawers” that are our corporate information management systems. For AI to deliver on its promise, these junk drawers need preliminary organizing structures before the vision of conversational interactions can be realized.

    In other words, you can’t have AI (artificial intelligence) without IA (Information Architecture)

    So what goes on behind the scenes to make this "magic" happen? Knowledge engineering helps businesses realize the promise of AI on enterprise search, information access and knowledge management, and allows us to progress toward establishing an intelligent workplace.

    Search as a Conversational Interface

    Search is a conversation. When you type a query into any search engine, it returns results that are an approximation of the searcher’s intent. 

    “Web queries are short, ambiguous and an approximation of the searcher’s real information need,” according to search experts Ryen W. White, Joemon M. Jose and Ian Ruthven. Therefore, disambiguation is typically part of the search interaction. (If you say, “Tell me about AI,” I will respond with a clarifying question “What would you like to know about AI?” to narrow the topic.) 

    Chat is a natural extension of the search interaction. Chat feels more natural with the advent of voice recognition and natural language inquiries, such as when we ask our smartphones where the closest gas station or sushi restaurant is located. It is the transformation of queries into natural language. 

    Early experiments in natural language queries have grappled with the challenge of ambiguity. Words have multiple meanings (i.e., “what is the stock price…” can result in finding the price of “a unit of financial value,” “the handle of a gun," “a base ingredient of soup,” etc.).  

    The marketing for Amazon's Echo promises users can search for a large scope of topics via the chat interface, Alexa. This broad approach does not align to workplace needs, where language and terminology is specialized. Generalized needs cannot be implied from the variations in users’ requests.  

    Corporate information is more nuanced and complex than gas stations and restaurant locations.

    Conversation Gets to the Heart of Search Faster

    Traditional search is limited because it does not rapidly filter the query to its true essence. People use search terms that fail to deliver the richer meaning that a conversation can elicit. 

    But when people talk, they disambiguate. They have a conversation which gets to the heart of the query faster. This process occurs through a series of refinements that come as a result of the give and take that questions and answers provide. For example: 

    CFO: “How can I help you?”
    User: “I need the quarterly report numbers for net revenue”
    CFO: “For what division”?
    User: “Consulting services”
    CFO: “For which Fiscal quarter?”
    User: “Q3 and Q4 2016”
    CFO: “Here is a link to the 2016 annual report. Page 16 has Consulting Services revenue by quarter. Does that meet your need?”
    User: “Yes, thanks.”

    The conversation zeroes in on the specific need through this refinement process, and saves the user the trouble of looking through a long list of search results and opening individual links repeatedly until they find the information they need.

    An iterative exploration of a topic through conversation provides a higher chance of achieving a relevant answer, as opposed to receiving an entire set of possible answers at once, as in an FAQ setting or a Google search result. 

    This is what AI brings to enterprise search. The iterative process results in higher efficiency (time-to-information), increased accuracy (retrieval of the right information), and a greater utilization of corporate information assets (information is put to work). 

    What AI-Driven Search Looks Like in Practice

    AI-driven search, in the form of digital assistants and chatbots, builds on activities we already carry out today. For example, a popular corporate travel administration portal includes a chat box where users can enter free-form text describing their travel needs:  

    "I need to fly from Boston to Chicago on March 20, and return on March 24."

    The system parses the request, conducts the search and returns the same flight choice result screen that is produced using the standard form-based interface. The difference is the user made the request in conversational language, avoiding the confusion that can occur when switching modes to accommodate the browser-based flight search form fields.

    Here is a more mundane yet very useful example of AI’s potential impact as a new search interface that you can try today: 

    Ask Siri, “Show me emails received from Joe Smith since Monday.” 

    Assuming that a user has been performing tasks on a Siri-enabled device, this capability can transform a very common business action — searching for the latest version of a slide deck or proposal — by searching for the colleague who most recently sent it to you. 

    Test this form of search with your smart phone and see what you get — or your Mac, where the latest version of the OS now supports Siri as well.

    Imagine what else we could do as these assistants and chatbots are refined. 

    Then There's the Rub

    While this simple example represents the potential for a tangible result, there’s always “the rub” — the very real set of challenges and hard work that brings the potential from hype-driven promise to business reality. 

    In the example above, the data is structured. Emails have a send date and a field containing the sender. Underlying the promise of AI-driven search is the requirement that businesses organize their information in a way that enables machines to parse queries in relationship to the structure. This is where knowledge management enters the equation.

    Tomorrow we will lay out the specific steps of this foundational work, providing a detailed explanation of the three major steps required to organize information so intelligent applications can consume it. 

    Editor's Note: This is the first in a three-part series. Read part two, on how to lay the foundation for AI and part three, on the challenges businesses face when introducing AI into the enterprise


    This series originally appeared on

    Earley Information Science Team
    Earley Information Science Team
    We're passionate about enterprise data and love discussing industry knowledge, best practices, and insights. We look forward to hearing from you! Comment below to join the conversation.

    Recent Posts

    [Earley AI Podcast] Episode 41: Ian Hook

    Ian Hook on Advancing Operational Excellence with AI and Knowledge Management - The Earley AI Podcast with Seth Earley - Episode #041 Guest: Ian Hook

    [Earley AI Podcast] Episode 40: Marc Pickren

    Search Optimization, Competitive Advantage, and Balancing Privacy in an AI-Powered Future - Marc Pickren - The Earley AI Podcast with Seth Earley - Episode #040 Guest: Marc Pickren

    [RECORDED] Product Data Mastery - Reducing Returns to Increase Margin Through Better Product Data

    Improving product data quality will inevitably increase your sales. However, there are other benefits (beyond improved revenue) from investing in product data to sustain your margins while lowering costs. One poorly understood benefit of having complete, accurate, consistent product data is the reduction in costs of product returns. Managing logistics and resources needed to process returns, as well as the reduction in margins based on the costs of re-packaging or disposing of returned products, are getting more attention and analysis than in previous years. This is a B2C and a B2B issue, and keeping more of your already-sold product in your customer’s hands will lower costs and increase margins at a fraction of the cost of building new market share. This webinar will discuss how EIS can assist in all aspects of product data including increasing revenue and reducing the costs of returns. We will discuss how to frame the data problems and solutions tied to product returns, and ways to implement scalable and durable changes to improve margins and increase revenue.