How bots can assist the core banking | equensWorldline
Blog / Innovation /

How bots can assist the core banking

Paul Jennekens

Manager Marketing

12 November 2020

How bots can assist the core banking

Unburdening bank IT through Conversational Banking platform solutions

Financial institutions are not only under pressure from the digitalized competition. Customers are also demanding services and forms of communication in real-time via different channels more and more. According to Veronica Alava, Business Developer Digital Services at equensWorldline, Conversational Banking supports the digitalization and automation of customer care without burdening the central infrastructures. Something that does not interfere with the core processes of a bank - apart from the provision of information.

Conversational Banking and chatbots are becoming increasingly important for the quality of financial services. Anyone who has a question about their finances wants to talk to their bank like they do for everything else in their digital daily life: via social media and messenger services. They also want to use intelligent voice control systems. As this becomes increasingly commonplace, account holders expect to avoid waiting times. Today, this is no longer restricted to purely administrative tasks. Also innovative services such as real-time payment transfers or personalized financial analyses are in demand. The security of data and the systems involved is tacitly assumed. The bank clerk, for his part, is hoping for relief through digital assistants who handle routine inquiries. In the case of complex problems, they are supported by an intelligent chatbot, which asks the customer the right questions and locates information or contact persons.

Future-proof dialog platforms

An intelligent text- and voice-based Conversational Banking solution meets the requirements mentioned above and acts as an interface to core banking connections. Powerful and future-proof systems are based on platform solutions that allow for a step-by-step and fast extension of the chatbot functions. On the one hand, they support the text-supported sending of messages via various channels such as the internet and mobile applications as well as social network messengers such as Facebook or WhatsApp. On the other hand, they enable voice-controlled dialog via speech recognition systems such as Amazon Echo, Siri, Google Home and Google Assistant. Statistics on the use of individual services also point out potential, bottlenecks and risks.

Crucial for the quality of the dialog and the resulting consulting is the targeted training of the chatbots for different concerns and conversation situations through a constantly improved Natural Language Processing (NLP). Engines that are specifically developed for processing natural language via voice or text use AI-supported technologies such as the Conversational-AI-Platform Rasa or Dialogflow to model dialogs. The goal is to better understand the wishes of a caller and to forward him/her to the appropriate department. In predefined conversation models all intentions are assigned to an entity. The machine learning platform uses score values to determine how confident it is in recognizing the customer's request. If this number is below the required threshold value, the assistant will attempt to gain further clarity by asking questions. As soon as the intention has been recognized with sufficient certainty, the request is handled in a relevant process - such as the execution of a bank transfer - or, if necessary, forwarded to a human clerk.

Connection to the Core IT

Financial institutions demand simple and seamless access to such AI services, but do not want to overburden or modify their existing core IT platforms and related infrastructures. Conversational Banking service providers such as equensWorldline provide a remedy: they take over the administration and backup of data generated by bots in their own environment. The bot microservices are provided by an internal open-shift platform and are located behind the service provider's load balancer and firewall. Connectivity and security are thus managed by the provider, and the actual IT infrastructure of the financial institution is not additionally burdened. The secure connection to the core banking system is established via VPN with a classic TLS connection. The data exchange takes place via standard connectors at the interfaces of the respective applications.

Platform providers ensure the security that is particularly important for financial transactions. Relevant are the general guidelines according to ISO 27002 with strict access controls, token-based security and peer-to-peer encryption of messages. Security can be tailored to the bank's needs with a mix of hot storage, where information is retrieved directly from the source in real-time, and cold storage for data that is less frequently needed and may have longer access times. However, the choice of storage concepts depends primarily on how versatile the conversation agents need to be.

It is important that the external platform of the NLP partner does not store any data that can be related to a person. Dialog processes can be traced in the follow-up. However, the NLP system only records anonymized data and is therefore not relevant for the requirements of the GDPR.

Banking of tomorrow

Conversational Banking has great potential for development. The improvement of natural language processing is an ongoing process. An important feature for the future is the recognition of emotions and moods. The goal here is to adapt bot communication to the behavior of the counterpart in the current interaction.

However, more complex areas of support or security cannot be left to the bots and their Artificial Intelligence in the future. This includes, for example, the detection of fraudulent mechanisms that require the creation of a file about the incident. Here, compliance guidelines and core banking IT infrastructures do not allow web-based services. However, the detection of unauthorized credit card use, for example, and the necessary escalation of such problems to the security department and the consultant can be improved by automated agents.

Practical example BNP Paribas

The major French bank BNP Paribas relies on text and speech recognition in all channels of customer contact. Account holders contact the bank via the website, social media such as Facebook Messenger and various devices. Bots answer classic questions about opening an account, the next branch or activating a digital key. Account balances and payment transactions can also be queried.

The French bank has developed its own solution for integrating Google Home and Google Assistant together with equensWorldline. New methods were used to model visual communication processes for both text-based and voice-supported dialog guidance for Google Assistant. It was important to understand exactly how the customer uses a particular channel in order to meet their expectations. In order to verify the models, simulations of conversations with employees were carried out. On this basis, interfaces between the respective applications were developed, which can be used for the development of further services. The topic of security was very important. The cooperation with equensWorldline was also characterized by a high degree of flexibility and transparency - for example by using project management software solutions such as JIRA or Confluence. The Voicebot has been in use since January 2020. BNP Paribas and equensWorldline are already working on the integration of additional functions.