Mobilizing a Nationwide Emergency Response to COVID-19

The Challenge

PicPay serves a rapidly growing customer base of 20 million customers and counting, the largest user base of any payment app in Brazil. But with the onset of the global coronavirus pandemic, the challenges of scale took on a new urgency.

The Brazilian government selected PicPay as the payment app through which all emergency aid funds would be paid out to families across Brazil. Through the PicPay app, users would be able to validate whether they qualified for aid with the Brazilian government and claim their benefits.

The logistical challenges of the plan were considerable: a spike in customer service contacts, a user base with limited digital literacy, and above all, a mandate to roll out a solution in weeks, not months.

To manage the expected volume and ensure all users were able to access their benefits regardless of tech savviness, PicPay designed a conversational interface: an AI assistant that led users through the entire process.

The Solution

With time of the essence, PicPay chose Rasa to provide the framework for conversational AI. The team’s past experience with Rasa made it a top contender during the evaluation process, and ultimately Rasa was selected over IBM Watson on the basis of both cost and accuracy. Another deciding factor was Rasa’s open source approach and the visibility it offered into constant improvements within the product, as well as the ability to support language localization in Portuguese.

PicPay’s team of developers, content editors, and data scientists worked closely with the Rasa Customer Success Engineering team to meet an ambitious launch deadline of just two weeks. 10 business days later, on May 13, the PicPay assistant was live in production.

The Results

In just the first hour, PicPay’s assistant handled 3,000 transactions, averaging 100,000 users per day during the first week. Traffic has remained high ever since, with the assistant handling between 20k and 65k interactions per day.

With a combination of technology and nimble development, PicPay was able to mobilize the distribution of aid across Brazil. The success of PicPay’s Rasa assistant has led the team to look toward future improvements, including integration with Zendesk and WhatsApp.

With the help of the Rasa team we managed to put an assistant in production and go live within 10 business days. Rasa is portable, scalable, and it’s very easy to develop a chatbot application on Rasa.

Isaac Ben-Akiva, Chief Data & AI Officer