All Posts

Putting Your Chatbots to Work

Don’t obsess about the technology – focus instead on equipping bots with carefully structured content and data to give customers what they want, sometimes even before they ask for it, say experts at Earley Information Science roundtable. 

Companies can be easily dazzled by their newest workers, the chatbots. They’re on the job 24/7, easing the load of “real” employees throughout the day and night. They never complain or get sick. And they don’t ask for raises. But companies can be just as easily disappointed when customers become frustrated by opaque and time-consuming interactions with the bots. 

It’s not the bots’ fault, said a panel of artificial intelligence experts at a recent Executive Roundtable hosted by Earley Information Science (EIS), a leading consulting firm focused on organizing information for business impact. 

Yes, chatbots are loaded with the bells and whistles of AI technology. But content and data are their lifeblood – and many companies are sabotaging their new armies of virtual workers by not supplying them with enough knowledge about products and processes to do their job. Too often, the companies turn to off-the-shelf software that misses the mark in packaging their information. What is needed, the experts said, are organizing principles and classification systems that are fully aligned with the unique aspects of each company’s business. Without those nuts and bolts in place, a chatbot can’t retrieve the information that a customer has asked for. Nor can it get to the next level –   anticipating unasked questions about other needs.    

“The data and the content are more important than the algorithm,” said Seth Earley, Founder and CEO of EIS. Notwithstanding all of its AI features, a chatbot is essentially “a channel to knowledge data and content, and when you have such a channel, you must have a knowledge architecture to support it.” Companies have to “repurpose” their existing content and data, he added, “to tag and organize it consistently so that it can be used for customer self-service, to support the call-center representative and across other types of AI systems.” 

The Feb. 21 discussion, “Knowledge Engineering, Knowledge Management and Chatbots,” was led by Earley, who was joined by Tobias Goebel, Senior Director of Emerging Technologies at Aspect Software, a customer-engagement technology and customer experience company; Joe Gelb, President of Zoomin Software, which creates content publishing platforms; and Alex Masycheff, Co-Founder and CEO of Intuillion Ltd., a content management and software development firm.  

The panelists discussed how knowledge engineering supports AI’s role in improving the customer’s experience, including advances in conversational search and information retrieval. Key to the effort, they said, is putting the right information foundation in place.    

“Chatbots can answer user’s questions, but so can other communications tools,” said Masycheff. What sets them apart is that they are also able “to infer questions not asked and answer them as well.” But to pull that off, he said, they “must be able to capture context and be aware of the domain” in which they operate. And to make them aware in that way, chatbots have to be “fed with an ontology, a structured model of everything we know about this domain.”  

Another challenge is to help chatbots serve up all that information by engaging customers in “conversational AI.” 

Because the chatbot’s user interface is so different from the traditional graphical user interfaces of web sites, mobile apps and desktop applications, you have to make sure that the bot delivers information in discrete bite-size pieces easily understood by the customer, according to Aspect’s Goebel. Chatbots are dialogue-based, not screen-based. “It’s about gradual information discovery,” he said, “question by question, message by message.”

There are great rewards in getting this right. On the flip side, though, there are great risks in getting it wrong.

“If the conversational flow is not effective, the technology will be disregarded and have an adverse commercial effect,” said Zoomin’s Gelb. “Anything less than excellent is worse than nothing at all.” 

The panelists also discussed project costs (you can start small), governance (you really need it) and returns on investment (short- and long-term).

The roundtable featured a real-time survey of the webinar attendees:  

  • 37% said their knowledge management function was aware and another 37% said it was competent. On the high end, 18% called it synchronized and 3% choreographed. But 5% said their function was unpredictable. 
  • Bot projects are generally valued: they are a major priority drawing executive attention and funding (35%), they are in development or deployed (33%), or they are in the initial investigation stage (38%). Some projects have limited proof of concept (28%), but only 3% of the attendees said that their leadership doesn’t see the value.    
  • Which bots are under consideration? Support and customer service (66%), internal HR functions (43%), conversational commerce (37%), lead conversion (29%), other types (37%).  

Please use these links to access the roundtable webcast and a related article, “Knowledge Engineering: Structuring Content for AI.” Seth Earley will be talking more about knowledge engineering’s critical role in enabling chatbots and other AI-driven initiatives at the 2018 ACE (Aspect Customer Experience) event in Las Vegas, in late April. 
The Earley Executive Roundtable is an educational webinar series focusing on topics of interest in the areas of digital transformation and information science. Each month, EIS leads a lively discussion with a panel of industry experts.

The next roundtable is scheduled for Wednesday, March 28, at 1 p.m. ET, on the topic of “Product Data: The Great Enabler of IoT, Marketplaces and More.” To sign up, register here.

About Earley Information Science: EIS is a specialized information agency. We support business outcomes by organizing your data – making it findable, usable and valuable. Our proven methodologies are designed specifically to address product data, content assets, customer data and corporate knowledge bases. We deliver governance-driven solutions that scale and adapt to your business as it grows. For more information, visit