One of the greatest challenges central banks face lies in communicating with the general public. While many have begun simplifying their language to reach a broader audience, some have turned to new technology.
In 2017, the Central Bank of the Philippines (BSP) partnered with the Regtech for Regulators Accelerator to develop a ‘chatbot’ to handle complaints from the general public. The central bank wanted to develop a prototype that would act as an automated dashboard for customer complaints, and a contract was awarded to Sinitic, a company that develops multilingual chatbots.
Using a combination of artificial intelligence (AI) and natural language processing (NLP), the resulting platform acts as an addition to the BSP’s customer assistance mechanism (CAM). The central bank hopes the chatbot will provide a wealth of data and intelligence that can be used to understand consumer sentiment and identify emerging risks and opportunities.
The BSP also hopes this information will feed into the central bank’s supervisory process, ultimately strengthening trust in the financial system. “The chatbot, in effect, amplifies the voice of the financial consumer as input to policy-making, regulation, and supervision,” says Pia Bernadette Roman Tayag, director of the Inclusive Finance Advocacy Department.
Currently, CAM receives around 10,000 complaints and enquiries per year through email, phone and visits to the central bank itself. Processing these complaints has been labour-intensive; the central bank previously used an Excel-based data management system to track and handle each query, which resulted in longer-than-expected turnaround times. There was also potential for human error in data management as staff focused on processing rather than analytics.
The chatbot, meanwhile, has a more streamlined architecture for registering, classifying, sorting and analysing consumer complaints. Unlike other dashboards, Sinitic’s chatbot connects to messaging apps, websites and SMS. This allows consumers to lodge complaints through ‘smart channels’ such as Facebook Messenger and webchat. Complaints can also be lodged in English or Tagalog – the Philippines’ national language.
AI allows the chatbot to determine complainants’ ‘intent’ and correctly classify them into categories. The more complaints the chatbot answers, the more fluid and accurate responses will become, as it gathers more intents and complaints for its database.
The BSP has integrated a case manager interface that allows staff to view analytics in real time, and use these as inputs for evidence-based policy-making and supervisory action. “More advanced analytics are programmed for addition to the chatbot in the future,” says Tayag.
By leveraging AI and NLP for its CAM, the BSP is a leader among central banks in using regtech to improve communication and supervisory processes.