Why Enterprises Are Using AI And Chatbots In Learning

Why Enterprises Are Using AI And Chatbots In Learning
Summary: Moving on from microlearning to conversational learning and engagement.

Learning Conversations With Persona-Based Chatbots

We first created our chatbots for learning more than 2 years ago, to experiment with the pedagogical impact on learning. What we found was astonishing.

Our Own Experience Creating Chatbots For Learning

To our surprise, the learner ratings shot through the roof, with an average of 4.5 to 4.6 ratings out of 5.0 for online learning. These ratings would have been extremely commendable for classroom training and with online learning, typical ratings are usually much lower. With our newly created chatbots for learning and Artificial Intelligence (AI), we also won 2 national tech-enabled learning innovation competitions in Singapore at the national level (in 2017 and 2018).

The message that we received was the need for users to see eLearning as interaction with human experts rather than with books and chatbots plugged that gap with persona-based chatbots. This conversational approach made learning fun, less formal, more timely and customized. It also modeled after social messaging apps which lowered the mental barrier to learning.

We will describe some of our own experiences in getting different enterprises on board to use some technology in their current work processes. A lot of the technology is new and still evolving, we will have to learn, adapt and keep moving as more options become available.

Let’s start small but strategic.

Advantages To Using AI And Chatbots

1. Persona-Based

One fun and important feature that a chatbot for learning supersedes online learning resources is that you are talking to an eCoach rather than consuming an eBook. The former is active and responsive. The latter is passive and static. Hence, we have created chatbots which have encouraged learners to go have their meals first before taking on this course. The chatbot can also gently chide learners for not remembering concepts or getting answers wrong after so many attempts. The localization possibilities are immense, and you can play the fun colloquial card with language nuances incorporated. This can be a mildly responsive avatar that looks at you each time you click or input a response. It would just be fun to play with.

Imagine if you have a fun workplace chatbot that can inform you of the latest staff news and product information for quick learning off your mobile phone. Wouldn’t that be awesome?

2. Conversational Learning Design

The WhatsApp style of passing nuggetized content to learners makes it more palatable. For example, there is an onboarding chatbot putting out a video introduction by the CEO.

This makes the learning—literally—bite-sized. Staff can be asked to give quick comments on what they have seen and these comments can be sent to a real person to follow up if necessary.

Good conversational design is critical to make the discussions authentic and informative. Learners are given opportunities to respond and check their learning. The pedagogy is coaching-based and facilitative. It builds on microlearning design with short learning loops spiraling one over another. At the same time, conversational design makes the learning fun and authentic, in as much as how we learn from someone more experienced than us. We engage in meaningful conversations. Hence, it is not just about microlearning. They are different experientially and pedagogically. What do I mean?

4 Features Of Conversational Designs

Firstly, with conversations, the interaction among the parties is co-constructed. The exchanges do follow the typical social rules of conversational engagement. As most people have experience in conversing with others whether face-to-face or via social media and messaging apps (e.g. WhatsApp), this creates a low-barrier entry into tech-enabled learning for most people. With a WhatsApp conversation, people may be less inhibited to enter into a learning conversation compared with entering a Learning Management System with unfamiliar buttons and formal rules of engagement. Conversational learning is akin to entering a room with an eCoach compared with microlearning, where learners enter a room with a book on the table.

Conversational learning empowers and respects the learner as a contributing partner to the learning process. Learners are expected to propose new ideas and refute points which they do not agree with. Learners can leave conversations or enter new branches of the conversation based on their preferences in a very natural manner.

For example, with one of our chatbots on customer service training, the chatbot is named Aunty Judy, after an actual trainer in the organization. She is well-liked and is always concerned about the well-being of learners. Here, we crafted her to ask if learners have eaten and if they have not, to go and eat first. I doubt many of us have encountered eLearning resources that ask us to go for our meals first.

Secondly, with conversational learning, branching is an innate part of the process. It is learner-led not just learner-centric especially if the learner is somewhat competently skilled. By allowing learners to select their learning pathways, learners have the autonomy to drive their learning outcomes, from the content and experience perspectives.

You can see this quite clearly from this figure above on the motivations for the learner to get into this chatbot for learning. Following which, it is simple to channel the learner into a different branch if needed. The branching can be invisible to the learner and subtly carried out if needed.

Thirdly, the conversations can be fluid and expressive, depending on what the learner wants to say. For example, we ask learners to reflect or comment on specific content that they have just seen.

We have also asked the learners to share what they think about the organization after going through a part of the onboarding process.

Fourthly, you can inject some machine intelligence into the chatbot for learning. Either keywords or a link to the Natural Language Processing model would help the chatbot interpret open texts. For example, there is a chatbot responding to ‘of course’ as an alternative to ‘Yes’ and continues the conversation from there. It makes conversational design more natural and meaningful. We have also developed chatbots that engage with potential customers by asking questions which are direct and informal. This leads to relationship building with the customers especially if the persona is clear and strong.

Finally, the engagements with the learners can be automated according to certain timelines. For example, in an orientation chatbot that we are working on for a university, there are specific moments when we want to engage the undergraduate (e.g. 2 weeks before orientation, day of orientation and 2 weeks after, etc.) to send critical pieces of information and to collect useful data about their needs. Some students may be lost on the campus while others may have problems collecting their student passes. This allows the chatbot to specifically target these needs and provide customized support. I can imagine how relieved parents can be if the chatbot can notify them that their child has reached the campus safely and registered their attendance or results of critical examinations are also made available to mentors of these students at certain junctures to monitor the students’ progress over the years of study at the university.

There are many other advantages to using chatbots for learning and engagement. It is critical to understand that our approach is not to open up the conversational exchanges too early but to structure the process so that the learner experience is specially and carefully curated to create a positive one from the start.


Going forward, we are experimenting with new ways of making User Engagement more personable and direct. Allowing experts to create their chatbots for learning and engagement is also something we are working on, with educational institutions and enterprises.

Using AI to help our learners learn faster makes sense if we want to compete with AI for jobs. The future of engagement and learning is here, and we need to work and learn fast; if not faster.