AI-powered Digital Assistants (DAs) have gained popularity over recent years. According to a study by Juniper Research, consumers are expected to interact with voice assistants on over 8.4 billion devices by 2024, surpassing the world’s population. Unlike chatbots, DAs have evolved over the past few years from rule-based responses to function as fully integrated end-to-end customer journey mapped solutions with integration flows.
The rule-based approach used to take the user through several predetermined steps to achieve their goal. With the introduction of pre-trained language models and advanced NLP algorithms, DAs were able to understand user requests better. These improvements led DAs to be conversational. DAs were able to evolve from understanding their users’ requests to being contextually aware of the user’s requests. Now, DAs can initiate conversations, understand the user’s request, be contextually aware, lead users through complex processes, integrate to external systems, communicate via speech commands, and efficiently complete user-requested actions from applications.
MillenniumIT ESP has deployed FinBot – the Banking DA – for financial institutions to enhance banks’ customer experience and approachability. FinBot can converse with a user while being contextually aware. It is capable of providing product information related to accounts, loans, credit cards, value-added services, general information, nearest branch/ATM information, check credit card balances, account balances, suspend misplaced cards, generate leads, handover the session to a live agent, raise a complaint, and many more features. Below are a few advantages of FinBot:
FinBot can understand human conversations and deliver a personalized and conversational experience to users. The DA improves customer experience by responding to customer inquiries in real-time and maintaining positive NPS and resolutions.
The Banking DA can assist customers and employees 24/7 via various channels such as WhatsApp, Facebook, Skype, MS Teams, Telegram, and Twilio, among others. It is reliable and can handle a number of requests simultaneously.
FinBot is able to handle customer inquiries from digital channels with minimum human intervention. This would enable bank agents to focus on more high priority, important tasks and attend to customer queries only when needed.
When DAs are incorporated, banks can understand their customers’ sentiments, most frequently inquired products and services and other insights to judge their areas of interest to use for targeted campaigns and promotions. Real-time insights will assist banks to take more strategic decisions based on concrete data.
Below are a few features and functionalities of the FinBot solution.
The DA understands the context of the user’s query without the user explicitly mentioning it. For instance, the user could ask “I want to know the nearest branch,” and the DA would respond with the nearest branch information based on the shared GPS location/entered city. If the user then asks, “Can you provide me the contact information,” the DA would know the user is asking for information on the nearest branch information shared and will provide the response accordingly. The DA is also designed and developed to understand the user’s query no matter how complex the command sentence is; it will understand all relevant entities mentioned by the user and provide the final response. Auto-correction and misspelt words are also supported.
The DA could trigger RPA processes. Information entered to the DA could be used to fill and submit external forms without manual intervention. The DA could trigger the first level verification RPA process of a customer based on the information provided. A user can type a message in the DA like “I want to disable my debit card,” the DA would then process the request and trigger the RPA bot. The RPA bot would connect with the on-prem legacy system, and once the task is completed, the RPA bot would pass the status of the response to the DA which would provide the relevant information to the user.
FinBot could be integrated with the backend systems of an organization. This includes integrating with CRM systems, submitting support tickets to ticketing systems and integrating with on-prem core-banking systems to get account balance and credit card balances. The high-level process of checking the credit card balance through the Banking DA is illustrated below.
Using the Banking DA, organizations could engage with their customers on multiple platforms. A DA could be embedded to the company website, Facebook Messenger, Telegram, WhatsApp, Skype, Twilio, Slack, and MS Teams, among others.
When a customer prefers to speak to a live agent, the DA could hand over the session to the live agent with the customer conversation history. This ensures a smooth transition for the customer and makes it an easier process for the customer care agent to take over the conversation.
Customized UI components, including rich user controls such as cards, forms, carousels, menus, multimedia content, and customized branding can be incorporated to improve the overall user experience.
DAs have the capability of conversing in multiple languages. At MillenniumIT ESP, we have deployed DAs that support English, Sinhala and Tamil languages to ensure that more people can access the DA in their native language.
While customer-facing DAs mainly support customer care agents, they also provide key metrics for managerial decision-making. Through the insights and analytics provided following the deployment of a DA, key metrics can be collected. An organization could identify their popular products, customer queries and expectations, identify seasonal changes and gather other key information for their decision-making needs.
Users are able to interact with voice DAs via voice commands. Custom Neural voice is also supported, which allows users to create a customized voice for their brand.
The DA consists of pre-trained banking and financial language models for account information, credit card information, loan information, value-added service information, e-banking information, and general information enabling the DA to be rapidly deployed.
Continuous re-training and improvements of the DA will be done on a daily basis to improve the customer experience.
FinBot is powered by Microsoft’s Bot Framework and the Azure Cognitive Services.
The Microsoft Bot Framework is used to build DAs that use natural language understanding, speech, questions and answers, etc. Microsoft Cognitive Services are used to build cognitive intelligence to the DA. Microsoft’s Language Understanding Model is used to predict the overall meaning and retrieve detailed information of the user’s conversation. QnA Maker is used to build a knowledge base by using question and answer pairs. Azure Storage is used to handle temporary session data. A site-to-site VPN is integrated to securely connect with on-prem systems.
Relevant security measures are undertaken to secure the banking DA by adhering to PCI Security Standards. Azure Bot Service is continually expanding its certification coverage. Currently, Azure Bot Service is certified with the following certificates. UI-level masking and encryption mechanisms are incorporated to ensure that all sensitive information exchanged are secure.
MillenniumIT ESP provides end-to-end DA solutions. Each solution maps the customer journey to provide a seamless user experience. Development best practices and processes are followed to create production-ready DAs.
Our FinBot solution is also recognized as a preferred solution on the Microsoft Marketplace and it is Microsoft co-sell ready containing pre-trained models for banking and financial institutions.
With people increasingly adopting digital platforms and virtual consumer behavior practices normalizing across industries, DA’s will soon become the preferred choice for customer engagement. These versatile artificial intelligence applications with multiple industry flexibility will liberate individuals from mundane tasks and help humans focus their energy on more creative endeavors.