For many banks, launching a virtual assistant might be the first time they actually engage with their customers in open-ended conversations. In many ways, it’s like giving customers a blank text box or open mic and asking for their true “voice of the customer” – in real-time and across channels. Responding to all of the idiosyncratic ways people communicate puts all virtual assistants through their paces, as well as the teams that manage them – and that’s why it’s important to make sure you have a system that can rise to these challenges and exceed your customers’ expectations.
These virtual assistants and bots are designed, integrated, and deployed with conversational AI platforms. Whether a bank chooses to build their conversational AI systems internally à la Bank of America’s Erica or license a conversational AI platform with built-in deep financial expertise like Wells Fargo, TD Bank, J. P. Morgan, and DBS, it’s imperative that the executives and employees across functions are aligned about what it takes to deploy and maintain a virtual assistant. Some of the world’s largest banks are years into successful AI deployments across many platforms and languages.
In fact, large banks are leading the AI revolution in financial services – nearly half (48%) of banks with over $50 billion in assets have already deployed AI solutions, compared to just 7% of banks with $1-10 billion in assets.
Ensuring your bank is ready for conversational AI is a major consideration. Deploying virtual assistants requires a myriad of functions at a bank to align – product lines, tech teams, customer care, compliance, and channels to name a few. Together, they will navigate a number of stages – from designing, integrating and piloting, to deploying, improving and expanding the virtual assistant to new products, channels, and markets.
Every bank should assess their organizational readiness and carefully consider these factors.
Integrating data and infrastructure
Access to real-time data is crucial for creating personalized and contextual conversations with your virtual assistant or bot. In order to design conversational AI experiences, first determine your most frequent, high-value use cases and then work backward to identify which data is needed to feed the platform to enable successful interactions. You will want to assess the quality of that data and determine whether it should be augmented to support the conversational experience you would like to provide. How is that data structured? Where does it reside? What are the points of integration to your back-end? These should all be known and understood before you start designing your use cases.
Different use cases require different data sets. For instance, a seemingly simple query like, “How much money do I have?” requires orchestration across all bank accounts, not just a checking account. On the other hand, a query like “How much did I spend on my trip to Vegas?” requires data to be augmented with geographical information. Regardless of your use cases, you will likely need to structure and augment data to meet your objectives. Fortunately, you can leverage tools you already use for enhancing and cleansing data to prepare your data for conversational AI.
Beyond core banking data, identify the other systems within your bank that can add value and enhance the customer experience via integrations with the conversational Al system. Live chat integration, for instance, can serve as a fail-safe for any interactions that require human intervention.
It can also be triggered strategically; if a premium customer initiates a conversation with your virtual assistant, you could set triggers that direct the virtual assistant to connect the premium customer directly with an agent for a “VIP” experience.
Integrating CRM systems can also help paint a more holistic picture of the customer and provide proactive engagement opportunities. A customized offer via virtual assistant can feel more personal than receiving the same offer via email or direct mail.
Creating content for entirely new customer experience
Once the pipes are in place, you need to develop the strategy for the content that’s going to flow through to your virtual assistant. Assess your bank’s existing content and determine the best sources to integrate into the platform. Many banks leverage the content at their agents’ fingertips, such as customer support materials, as well as website FAQs.
However, keep in mind that a carbon copy won’t work. This content needs to be shaped and refined for conversational interactions. That means short, bite-size replies – paragraph-long answers that are appropriate on your website are out of place in a conversation with a virtual assistant.
Since the virtual assistant is an extension of your bank, it should also mirror the personality of your brand. When customers ask CapitalOne’s bot Eno, “What’s in your wallet?” it replies “I don’t have a wallet, but if I did, it would just have pictures of cats.” What your bot says is the crucial ingredient but how it engages can contribute to a delightful customer experience.
Securing early cross-functional collaboration and approval
Without buy-in from security, compliance, and legal, your virtual assistant will never see the light of day. It’s critical to involve these key stakeholders as early as possible. Conversational AI is a new breed of banking experience, and the sooner you educate and inform these teams, the smoother the process will go. If you opt to license a platform with finance industry expertise, their team should be well-versed and experienced in explaining how their platform has mechanisms in place to meet your stakeholder requirements, such as encrypted back-end communications to protect customers’ personally identifiable information (PII).
For content approval, be sure to bake extra time into this process with your legal and compliance teams in particular. They are accustomed to approving every word of content and will need to adjust to the intricacies of conversational AI. You simply cannot provide exactly everything the virtual assistant will “say” since many interactions will be machine-generated based on contextual situations. Educating these teams early about the inherent differences of a conversational AI experience will help you avoid a potential delay-inducing stalemate down the road.
Designing deployment plans for GTM and future
Deployment is another key consideration. While banks are early adopters of AI solutions, they lag further behind in cloud adoption due in part to security concerns. Deployment models run the gamut – pure cloud, on-premises and hybrid being the most common environments among banks. It’s important to have a well-defined deployment strategy that your conversational AI platform supports and has experience implementing. Your solution provider should conduct a thorough evaluation of your organizational model and be able to support and implement your specific architecture standards and requirements.
Confining conversational AI to solely on-premise environments carries some disadvantages. One shortcoming is that banks do not benefit from instant AI training improvements as a result of ongoing learnings from customer interactions. Consequently, it’s becoming common to take a hybrid approach where production infrastructure is on-premise and processes like AI-training and analytics are performed in the cloud. Setting the right foundation is critical to your virtual assistant’s ability to engage 24/7 with customers at scale across multiple channels (e.g., website, mobile app, messaging platforms) and roll out new services in a streamlined manner.
Getting conversational AI off the ground is no small feat, but it is worth the sweat equity. Just ask digibank, DBS Bank’s digital bank, whose virtual assistant handles 82% of all inbound customer requests without ever needing to involve a customer support agent. With the right operational and technical readiness levels, you too can start reaping the rewards of conversational AI.