Categories
Digital bank Fin-tech

The 9 steps to build a great banking Virtual Assistant

Globally, many sector leaders have already launched their chatbots or virtual assistants and enabled their customers to experience the conversational interface. Many others prioritized this new channel on their digitalization agenda. The banking customers can experience a very large spectrum of chatbots from the ones with very basic functionality and capability to the sophisticated virtual assistants. Apparently, ”conversational banking” is the new digital framework of banking, although it is in the initial phase.

In this blog, I want to focus on the basic points that banking executives should consider to provide outstanding conversational banking experiences to their customers. And I talked to Mete Aktas, the CEO of Cbot, a conversational AI company that has developed AI-based banking chatbots for the customers of the leading players of the sector.

He summarized the key points for successful banking chatbot by 9 clear principles

It is essential to consider chatbot business as a part of larger AI strategy rather than a simple subsidiary channel. So banks need to have a long term view with a strategic perspective regarding their banking chatbots. With its broader terminology – “conversational banking” is a huge trend that defines the future of digital banking and it deserves to invest in.

Before starting, the banks should clarify the aim of adopting the chatbot, the problems they will address, the key performance indicators of it and the long term phases of their conversational banking agenda. Banks often choose to start with a limited function and extend it gradually – this is much more user-friendly than changing everything at once. Also, in most cases, they prefer a soft launch that is followed by a full public launch after improving the chatbot by the learnings within a beta period. Mete Aktas adds that Banks may start with a limited area, but they should keep in mind that they need to provide a comprehensive experience through this channel in the mid-long term. They must have a very clear idea of where to start and where to go.”

Banking service is one of the most sophisticated ones within the service sector – utilizing a mortgage is definitely more complex than ordering pizza. So banking chatbots developed through an advanced NLP (natural language processing) and a machine learning technology can deliver a more leveraged customer experience, whereas the rule-based ones are limited to satisfy the high customer expectations. The AI-based chatbots can understand the natural human phrases, misspellings, synonyms, nuances in a language and provide the relevant answer or action. In addition, contextual chatbots can go beyond understanding the human phrases and track the context of the dialogue to maintain it from the beginning to the end. So, AI-based chatbots are more suitable to the banking service in terms of the value that they create for the customer.

Training the chatbots with industry specific data set is the essential part of this business. The banking chatbots definitely must be trained with the deep financial expertise in order to be fluent in communicating with the banking customers. Choosing a pre-built banking chatbot instead of training an industry ignorant one absolutely minimizes the cost & investment and accelerates the time-to-market for the banks. Mete Aktas, says “The ideal banking chatbot should require no training cost for the bank and should be pre-trained like a 5-6 year experienced banking agent.”

Actually, the each type of banking chatbots –  informational, transactional and advisory –  require a different set of training data.

Aktas, continues “Contrary to the common belief, informational use cases require larger number of intents, or the ability to turn written phrases into queries than the transactional and advisory use cases.”  Whatever the type of the banking chatbot, it is important to deploy it as already-trained with the sophisticated and comprehensive data set by the vendor to minimize the training cost of the bank.

Today’s customer expects the same level of fast and frictionless service from every sector. Conversational banking has the potential to go beyond the current mechanical and static interfaces like web and apps, and deliver a faster and more humanized experience. So it has a big potential in terms of experience. But, for banking chatbots designing the experience is a very tough task in which each and every detail should be considered. Maintaining a dialogue end to end seamlessly is the most important part of the customer experience. The dialog flows should be designed with a proper CX perspective. Additionally, calculations, graphs, carousels, maps, visuals enrich the experience of a banking chatbot. Avoiding complex expressions and keeping the language simple, injecting humor and using emojis are some common ways to delight the customers. Another aspect is personalization – in retail banking, chatbots should offer a more convenient, personalized and enjoyable customer experience. Human hand-off when necessary with a seamless transition is another aspect that carries the experience to a higher level.

As the financial services industry has complex systems and workflows, it is crucial to design a chatbot framework that takes this complexity into account. Banks must integrate chatbots with multiple back-end systems and other sources of information and ensure their big data and conversational interfaces work together. Because a banking chatbot that is smoothly integrated with the infrastructure of the bank can deliver a personalized experience to the customers. At the same time, a banking chatbot can also collect more data to be analyzed for more personalization. This smooth interaction and ability to work in coherence is not an easy task both for the bank and the chatbot vendor, but it is important to create a really “living” chatbot within a bank. 

Banks also may prefer to put more than one chatbot behind a conversational interface, then the maintenance and orchestration of multiple chatbots gains importance. Some vendors are capable of being a master assistant, routing requests to several back-end chatbots covering different domains. This orchestration requires an NLP engine and a middleware layer. Another important aspect is the capabilities of the vendor to build the chatbot both on-premises and on-cloud. The most competent ones can offer both of them, or a hybrid installation model. 

Integration to the internal systems of the bank is not sufficient for a comprehensive banking chatbot. Its ability to help the customers whenever and wherever they need is also important. Besides banks’ own delivery channels like website and mobile app, messaging platforms like Facebook Messenger, Whatsapp, WeChat or voice assistants like Google Assistant, Alexa, Siri are hubs where banks should consider to integrate. As today’s customers want to receive services without switching among different platforms, banking services will also be available at the new digital hubs as far as the regulations allow.

Building a chatbot should not be considered as a one-shut project, but rather part of a larger strategic initiative. So a successful banking chatbot delivery requires a long term approach and commitment for both the bank and the chatbot vendor. The banking chatbot business requires an ongoing process of training, support and improvement. Defining poorly performing areas and improving them continuously to increase the accuracy and the quality of experience is an important part of the chatbot service. It is the vendor who should monitor the performance and continue its support and improvements for as an ongoing process, ensuring that minimum training and maintenance cost for the bank. Mete Aktas, from Cbot,  said “You should immediately answer a question regarding a campaign that is launched today. No banking chatbot can have an excuse like – “I am not trained about that campaign”. It will definitely diminish the quality of the experience and erase the trust of the customer.”.

Taking into consideration the wide range of possible functionalities of chatbots, security challenges cannot be ignored. First, banks should define a robust, role-based compliance and security framework that covers data storage & protection, handling and transparency. Second, bots are able to mimic human conversations, so social engineering attacks should not be ignored. Technological methods, including biometrics, two-factor authentication, behavior analytics and data encryption, can all contribute to overcoming security issues. Because of the security challenges and regulations banks mostly prefer on-premise installments, so a vendor should be able to provide this kind of a solution as well.

The most essential part of building a successful chatbot is partnering with the right vendor because it is the vendor’s capabilities, technology, experience, approach that will ensure the requirements we have discussed. Each vendor has its own strengths and weaknesses and no vendor can claim to cover all the needs of a bank, but our mentioned requirements are the basic ones and should be covered by an experienced enterprise vendor. Of course a vendor that best fits the specific requirements of the bank is the best one. Mete Aktas, discusses “In fact the banks are seeking partners that have the necessary experience and capability to collaborate with large enterprises in terms of both technical know-how, culture & approach. This asset has been a differentiator for my company for our bank partners. Because it is different to be a vendor and an enterprise vendor. And banks are very much aware of it”. Also, banks should be realistic, pragmatic and tactical when choosing a partner. Todays’s best partner may not be able to fulfil the requirements two years later, so an exit strategy should always be defined, the possibility of switching vendors should always be kept in mind.

Reference: https://www.forbes.com/sites/ilkerkoksal/2019/04/30/the-9-steps-to-build-a-great-banking-virtual-assistant/

contact@54.179.137.111.

+84-28-3535-7966.

Leave a Reply

Your email address will not be published. Required fields are marked *