Introduction
I waded into JRR Tolkien’s gargantuan epic ‘Lord of the Rings’ (LOTR) a little late in my life.
I have been guzzling books and stories in their varied forms since I was knee-high to a grasshopper, but I must confess, with LOTR, I found myself at a loss sometimes—with the characters, their history, their actions. According to the LOTR Project, there are a whopping 980 characters altogether if you consider Tolkien’s Hobbit, Lord of the Rings, and his posthumously published works.
To be honest, there were times when I wish I had a character map—something that showed each character’s history within the saga, when they first came in, who they interacted with, where they fit in the scheme of things—basically a Cliffs Notes for the characters. Sometimes it all got to be a little too much and I would be looking ahead to see how much of the book I’d covered and how much remained.
Then I thought back to all the J.T. Edson westerns that I used to read as a kid, and I remembered the detailed notes at the back of some of the books. Does the new character you just encountered in the story seem familiar? Refer to the endnotes and instant character bio and information recall.
It was cumbersome, having to flip back and forth between the story and the endnotes; which is why I was happy when I encountered some versions of Edson’s tales of the Wild West with these notes at the bottom of the pages as footnotes and not as end notes at the end of the book. All I had to do was look down toward the bottom of the page for a complete character sketch, without flipping a single page.
“How is this relevant?” you might ask.
If I were to draw a parallel, those humble footnotes in a pulp western novel were probably the first use case I came across of a just-in-time agent for query resolution.
Now consider this instead of the badlands of the lawless frontier, you are in an office where you are trying to get to grips with an application that is entirely new to you.
Instead of the exhilarating epic western, you are making your way through a curriculum consisting of 77 web-based training courses, designed to familiarize you with the new application and its workflows related to your role.
Each course has an average duration of 30-45 minutes.
Each course showcases processes to be completed in an application.
Each course has a plethora of technical terms and terminologies.
Now imagine you are going through this curriculum, putting in, say, an hour each day.
You have just started the 23rd course and encounter the terms “Business Partner ID” and “Global ID” that seem synonymous yet appear to have been used as if they mean two different things.
And you wonder…
What exactly is the difference?
Where does each term fit? When is one term used and when the other?
Where can I look to find out the difference between the two?
How can I do that without having to interrupt my course?
And the biggest question of all—Which of the preceding courses covered the difference?
Let’s kick it up a notch.
You are going through the demonstration for creating an entry for a new customer, and as you watch it, a fresh salvo of questions peppers your mind.
What if the customer changes their address in the future?
What is the procedure to update the address? What are the conditions to consider?
Where is that process covered in the curriculum?
In a classroom, there is always a facilitator to point you in the right direction. But what happens when you are doing the training as web-based training on your own—at your workstation, during your commute, or anywhere else? There is no guiding guru at hand to immediately answer any questions that come up.
How can we try and address the queries that the learners have while taking an online course, with minimum disruption to the learner and the minimum time away from the course?
In this article, we give an overview of Encora’s implementation of a virtual Intelligent Query Agent for query resolution, or chatbot, in an elearning program. This Query Agent was designed to reside within the course and respond to learners’ queries. We outline the requirements from the client, the solution that Encora came up with, and an overview of the tools, technologies, and framework used.
Requirement
A leading sustainable energy company in the United States reached out to Encora to redesign and build a digital learning program to educate their Customer Service Representatives (CSRs) on their Customer Relationship Management and Billing (CRM-B) application. The existing program was classroom and webinar based. The final redesigned digital learning program had 77 self-paced eLearning modules.
While conducting the program as classroom training over the years, the client observed that the learners often asked a lot of questions related to the program’s content. A few of these questions the client grouped together as FAQs. However, the client wished for a solution that would enable them to manage ad hoc, impromptu, and random learners’ questions in real time.
Solution
Encora proposed an “Intelligent Query Agent (aka chatbot)” to meet the client’s requirement of real time learners’ query resolution.
While developing the 77 eLearning modules for the digital program, the Encora team, in collaboration with the client’s SMEs, collated and documented all the questions that learners could possibly have about each of the modules. Client’s SMEs provided their input based on their years of first-hand experience in conducting the program as classroom sessions and webinars. Encora weighed in with its understanding of content and expertise in learner behavior and the science of learning.
The result was an extensive database of keywords, phrases, and frequently asked questions. This database formed the input that was used to ‘train’ the chatbot. The chatbot was integrated with each SCORM package and uploaded onto the LMS. As a result, when the learners launch any of the 77 course modules, the chatbot appears as a small clickable icon in the bottom left corner of the course window.
Anytime the learner has a question, all they have to do is click the chatbot icon. This opens a small popup within the course window and the chatbot welcomes the learner with a cheery greeting.
The learner can then type in their query or any keywords that they have queries or doubts about. The chatbot studies the query and responds instantaneously with an appropriate answer.
The chatbot facilitates the learning process by providing responses that help learners clarify their doubts, answer their unique queries, make connections, and actively engage with the content.
Going back to my LOTR experience, imagine if I had been able to interact with Gandalf in the story. I could talk to Gandalf, ask him for answers to the questions about the lore that bothered me as I read through the book.
Making of the Intelligent Query Agent (IQA)
Encora explored various options for developing the chatbot—creating a custom solution using open-source platforms like RASA, Amazon Lex, and Azure services to name a few. After a thorough comparison of the timeline, offered features, and cost, our experts decided to go with an Amazon Lex-based solution.
The AWS architecture presented a highly scalable and resilient cloud-based solution designed to meet the demands of the chatbot application while optimizing cost and performance. By leveraging various AWS services, the architecture ensured reliability, security, and flexibility in delivering the chatbot functionalities.
At the core of the Query Agent architecture is Amazon LEX, which provides the flexibility of Natural Language Understanding (NLU) enabling us to build a highly engaging, conversational chatbot.
To ensure secure access to the resources and protect against unauthorized access, AWS Identity and Access Management (IAM) is utilized for centralized access control and permission management. IAM enables fine-grained access control policies, allowing to grant specific permissions to resources based on their roles and responsibilities.
The Query Agent Web application has been developed by using web programming languages such as HTML, CSS, and JS. Subsequently, the Query Agent web application was embedded within each of the 77 SCORM packages which were then uploaded to the client’s Learning Management System (LMS) as eLearning modules for the learners.
Overall, the AWS architecture offered a robust and scalable foundation for the Query Agent application, leveraging the flexibility and scalability of cloud computing to meet the evolving needs of the business.
Conclusion
Query Agent or Chatbots have been quite a rage for some time. They hold a lot of potential and open up a world of possibilities. In a world where individuals are getting increasingly isolated, they promise to be everybody’s BFF.
eLearning can be a lonely undertaking with no cohorts, no guiding guru, and the only company is frequently the clacking of your keyboard keys and the clicking of your mouse button. Given this scenario, a query agent or a chatbot that can answer your queries not just in a dull clinical manner but in a conversational tone would be a welcome change. Especially if the chatbot in question is not just carrying on a conversation but actively aiding in the learning process.
It is much easier for anyone to talk to a chatbot without the fear of seeming stupid. With a chatbot, people don’t feel the weight of any judgmental eyes, real or imagined, burning into them.
Encora sees chatbots evolving into indispensable learning guides and assistants. They might not quite replace the proficient professor or the insightful instructor, at least not right away. However, we do see them being used more and more as learning advisers, shepherding learners through complex curriculae, steering them with tailored recommendations, providing real-time support, clarifying concepts, and answering queries. Integration of AI and analytics will further help with optimizing delivery and making the learning experience more engaging and dynamic—help bridge the gap between synchronous and asynchronous learning.
