Background

One third of people in Ireland say they aren’t confident in their digital skills. So we built GApp, a chatbot that helps digitally disadvantaged young people to build a CV, upskill with free digital learning tools and even apply for a job - all through Whatsapp.

I was responsible for the both website frontends and the chatbot. There were two frontend builds for this project - a CV generator and a form entry for a competition. Both were built using Angular. Users could enter their details into the form, adding or removing relevant education and work experience, and then generate a CV. The CV is generated within the users browser.

No data is sent or stored on any server.

The chatbot was built using AWS Lex. I have experience using Alexa which is built on top of Lex so I was very familiar in how to structure the chatbot and found it easy to get going. We utilised Twilio and the WhatsApp API. Data coming from WhatsApp was routed to our Lambda function where our intents would respond to the user via Twilio.

Challenges

Testing a chatbot is challenging. AWS does provide a way to test the chatbot within the Lex console, however doing full end-to-end testing is difficult. We had issues with Twilio and debugging where the issue was in the pipeline.

Handling script changes from copywriters and maintaining the chatbot flow can become tedious. To simply this, I created a flowchat so that both developers and copywriters could visualise the chatbot flow. This helped everyone stay on the same page and alleviate the issue of incremental script changes which had the potential to break the chatbot flow.