PAXI WhatsApp Bot

PAXI needed a solution that would improve their customers’ user experience and allow as much self-service as possible. Specifically, as a growing business, PAXI needed a platform to field general product questions (FAQs) as well as unique transaction related queries. The most common requests to PAXI’s call centre included parcel tracking and finding the closest PAXI delivery and collection point so the solution would need to interact with the PAXI API to return this information to customers. Where self service was not enough, the solution would also need to allow users to break out to a live call centre agent to help resolve their queries.

We assisted the internal product team to evaluate different solutions and refine key solution requirements such as cost, ease-of-use, customer support and ease of customisation.

We eventually settled on the following technologies for the solution:

• Twilio as the WhatsApp provider
• Dialogflow as the Natural Language Processing (NLP) technology and 
• Google Cloud Platform (GCP) to host the web services and related data.

Twilio was chosen because it provides a simple entry point to WhatsApp and allows developers to quickly create a highly scalable solution using it’s well thought-out and documented APIs.

Dialogflow was chosen as the NLP platform because of its ease of use and ability to be deployed across multiple channels with little change. As it also forms part of the Google product suite it integrates well with the various Google Cloud Platform (GCP) offerings we would be using on the back-end.

GCP was chosen as the back-end platform not only because of a number of specific client requirements but also because we found it to be very product and price competitive when compared with alternative offerings from the other major platform providers. A good example and a key advantage of GCP is the use of BigQuery to allow PEP’s internal Business Analysts to gain insight into customer needs by analysing the conversations customers are having with the bot.

The service was well received by users and within the first year of use we saw the number of queries logged on the service rise to over 100k a month and a significant drop in routine queries to the call centre. The solution continues to grow as we refine and build out the service to handle more query types.