Using AI Chatbots to Support Learning

Other Media’s Nick Oates was recently invited to Singapore by Hyper Island to build an AI-powered chatbot which would form the centrepiece of a learning day to support business transformation at Axiata (one of Asia’s largest telecommunications groups ).

The goal was to help 12,000 employees to think and act more digitally.

Hyper Island designs learning around making and doing. For this event a chatbot would play a central role in communicating with the participants, offering them interesting things to do, helping them to reflect and answering their questions.

Nick built the chatbot to run inside Facebook’s Workplace environment for business. This was a good choice from a user perspective because its similarity to Facebook made it very easy to use intuitively. However, it proved to be quite a difficult development environment because this version is still in its infancy. Documentation and support are still quite thin on the ground.

We used node.js to implement the software and the machine learning library from wit.ai (owned by Facebook) to understand what users were asking and create a natural conversational interface.

The biggest challenge was the short development time as the client had to run the event on a fixed date. With barely 8 weeks development time available our priority was to define the interaction model as quickly as possible and to implement a version that could be tested and refined as it was deployed.

We had 10 days of preliminary activities and engagement to get both the software systems and the participants ready. Then came Digital Jam Day itself. The whole organisation – a group of companies in 6 different Asian countries from Bangladesh to Indonesia – interacted together with DJ, our bot, at the heart of the communications.

It turned out to be a huge success!

Our biggest learnings are around the great potential for bots in supporting users in a wide range of activities including learning, job support and talent development plus the more commonly identified applications such as customer service and process automation.

One big surprise was how quickly participants got used to talking to the bot about a much wider range of topics than just the learning activities they were doing.

We are really looking forward to developing further chatbot and AI projects with our clients in 2018; working with more platforms in addition to Workplace and with more machine learning capabilities. If this sounds like it might be a valuable addition to your working environment please get in touch.