Conversational Ai: The Next Evolution in Banking

By Faisal Abbasi, Managing Director UK&I, IPsoft  

Automating business processes is a cost-efficient way to ensure smooth customer service. But there’s a substantial difference between quickly fixing a single problem, and consistently providing an optimal customer experience.

A personal banking concierge

With conversational Artificial Intelligence (AI), banks are starting to evolve customer service processes without having to invest in additional resource. It’s no longer just about automating the troubleshooting of tasks, or helping customers reset their passwords. In reality, the bank is giving every single customer a digital concierge that can help them find new products, research account and loyalty programme options, and even handle high-value tasks such as mortgage application processing or assisting the internal HR team with employee benefits selection.

Routine transactions that require manual completion by bank staff cost businesses 20 times more than transactions handled by customers online, according to Bain & Company. Rather than dedicating staff (and 20 times the resources) to handling rote processes, banks should be focusing workers on only the most important tasks. However, a critical distinction exists between what’s defined as automation and what qualifies as an AI-based solution. In other words: How you choose to automate your customer service tasks matters a great deal.

Chatbots won’t provide all the answers

When customers have a question pertaining to basic account changes, they will typically go on their banking app or to the website seeking an answer. Some of these questions may be simple: How do I transfer funds? How do I add my partner to my account? Companies often makes it easy for customers to search and find these answers within the existing content of their website. But what happens when customers are ready to take action on the basic information they’ve received? Many companies have adopted simple chatbots that can automate rules-based processes. If a customer asks Question A, the chatbot helps her complete Process A. This has served banks moderately well for the past five years. Unfortunately, if you’re one of the companies that has deployed a chatbot, your customers have likely encountered frustrating limitations that negatively impact user experience.

One of the critical drawbacks to working with a chatbot is its reliance on sequence to automate basic service needs. Customers may only visit a website with a single query, but they may also come to it with five distinct questions or think of additional things to ask during the course of their visit. Within the standard and scripted processes followed by chatbots, each question and its associated automation must be handled individually and within its own strict context. Deviation from or interruption within that context often leads to a dead end.

Faisal Abbasi

Automated multi-tasking

For example: A customer comes to a site and types to a chatbot, “I’d like to check my loyalty point balance, but I noticed someone just put a fraudulent charge on my account. I’d like to cancel my credit card and order a new one.” Within this single message there are five distinct business intents that must be automated. A chatbot that by design follows a strict sequence will not take into account any urgency to these requests (the fraudulent charge) and it will not find the most logical resolution path. Instead, a scripted chatbot will handle the first intent, and reports the customer’s loyalty point balance. Only then will the chatbot proceed to checking for a fraudulent charge. Clearly, the customer would view their priorities a bit differently.

Learning to prioritise

A digital colleague doesn’t need to accept information in a strict sequence in order to automate tasks. Although a digital colleague’s brain doesn’t entirely replicate human thought processes, it’s more nuanced than a scripted chatbot in how it processes information to make decisions – and it certainly provides more than automated processes. With the aforementioned fraudulent charge use case, a digital colleague will process all of the intents within the sentence (loyalty points, fraudulent charges, account cancellation and a new card request) and determine that pausing the customer’s account to prevent additional fraudulent charges is the obvious first step. Only after the digital colleague has handled the more important processes will it move onto simpler tasks, such as loyalty point balance checks.

The more complex the sequence of intents, the less capable a basic chatbot will be. This is particularly problematic when dealing with high-value tasks such as processing mortgage applications or helping customers book travel arrangements. If your conversational digital colleague is forced to follow strict sequences in order provide support, they will never be able to assist with and complete processes that require dozens or even hundreds of steps – some of which will need to be revisited multiple times within a single conversation – leading to undue time and effort on the part of a customer.

Don’t confuse automation with conversational AI. Basic automation delivered by chatbots makes simple processes easy for software to repeat, but that’s where the benefits end. Conversational AI from a digital colleague allows your business to automate at scale the nuanced, intelligent service provided by your best human employee, going ‘off-script’ to answer more complex queries and carry-out lengthy requests like mortgage applications. Ultimately, this means banks can deliver a positive customer experience for every single interaction, without the oft-encountered frustration seen when customers can’t get what they want from a chatbot. Banks that want to ensure consumer loyalty, higher customer satisfaction and continue driving a healthy bottom line should look to conversational AI as their next big initiative.