Fintech Chatbots
Hype or THE way forward?
Chatbots are one of the recent hot potatoes in fintech. We see bots popping up in bank announcements, on fintech sites and conferences, but what is their actual state and future potential?
History and present
Bots made their debut on their hype curve with headlines like „The solution that can replace customer engagement” and „Fully automated banks are the future!”. While there is certainly truth in these statements, bots are still lagging behind these expectations. If we want to be honest, currently even the best of them are only one trick ponies, dealing with one of the following use cases:
#1 The majority of bots are dedicated to speeding up the FAQ type customer queries by sorting and selecting the best answers to simple questions like “What should I do if I lost my card?”. It replaces the simple (and mostly underperforming) search-boxes and deflects callcenter incoming calls (a bit).
#2 There is a second, significantly more complex case when the bots are built into closed, authenticated environments to help the customers as their private assistant. Though mostly capable of giving answers to the previously mentioned FAQ questions, their real purpose is to create personalised replies and insights for their customers, proactively assist and execute transactions.
Despite their huge potential currently both of these groups lack the ability to handle complicated dialogues. When switching from simple queries like: “How can I save money?” (where multiple answer is can be a ‘good enough’ answer) to more sophisticated ones like: “Is there a good way to save money if I’m not able to set aside fix amounts each month?”, most bots fail to answer it properly. They are able to give an answer as they understand the intent (to understand saving options), but they struggle to understand the deepness of the sentence.
Most vendors believe the technology is still in its infancy and with time and increasing amount of user data bots will significantly get better. Although we all know that first impression truly determines the success or failure of the digital platforms: customers expect it to be perfect, otherwise banks will have a really hard time changing their view later on.
Our take on the topic
We believe that the case is similar to the generic AI. Everybody expects to become reality immediately, however this won’t happen from one day to the other, rather there will be areas that are already available like image recognition, natural language understanding, predictive analytics and recently machine learning. We must be able to carve out a realistic evolutionary path for bots as well, so for us the big challenge is to define how to get there step by step.
However it is obvious for the industry that it will take years to reach a human-like (natural) level of conversation with AIs, an average customer is far less aware of it. Let’s take Siri as an example: however iOS users are quite accepting of new Apple functions, their majority still pass on Siri despite the efforts made to make it better. This is the big challenge we see in banking today as well: the ability to create a good-enough experience for the first time to make your customers stick to the technology on the long run. We at Finatsra Labs have found a solution.
Shift from being reactive to provide proactivity
Analysing all the available data, digging deep into desktop researches, having debates with top domain experts all around the world and concluding several user test sessions with our solutions enlightened the way forward for us.
All journeys can be separated clearly into two major categories, and we take on those one by one
Most common reactive use cases banks may encounter like: What is my balance? / Is there a bill coming up? / Do I have enough money until the end of the month? These cases fall into the group that we handle with the common bot tools: text to speech, natural language understanding, free text answers, intent analysis, etc.
The second, much more exciting and challenging group covers the proactive use cases. As the name says, instead of waiting for the user to ask something the bot will proactively start the conversation. Although it may sound pushy at first, we found that by following a few important rules most customers don’t feel it intrusive rather the wow moments in their financial lives:
#1 Always start with something useful: a personal insight, a quick tip, a useful reminder. This is a nice way to convince the customers about the bot’s value. As a result, the response rate of bot messages becomes significantly higher than that of any other channel’s (including callcenter calls)
#2 Always ask only one simple thing at a time. It is vital to design every step of the conversation to be super easy. Even if this results in significantly longer conversations customers feel less frustrated all along. We already understand that Microcopy can be considered as one of the most important areas of digital engagement, and will be even more critical in the no ui world
#3 You can use this channel for sales purposes, but never design a chat-flow where ‘paying’ is the only option. This experience derives from the freemium model: people tend to accept purchase related options if they feel it is a free choice, and have to do it only when they already the understand the benefits of their decisions
#4 There is no one-size-fits-all approach. Banks have to understand their users, be able to group them and come up with individual plans to tailor the schedule, topic and wording according to the preferences of the given groups. While a gen-z member may find gifs and smiles natural but an affluent customer whose fortune is held by the institution may find them preposterous.
Guided dialogues above all
Although proactivity is a good trick to increase the usage of banking bots in a way that the bot is always conscious about the topic, it’s still not solving the ‘being able to give the right answer’ problem. To solve this, we looked into conversation-based mechanisms and discovered that the gaming industry faced the same problem many years ago when they wanted to tell long stories in an interactive way.
We started to implement ‘guided’ conversations similar to choose your own adventure games/books where you as a player are able to select from different ‘options’ to move forward in the story. We realised that setting up a complicated saving plan, assisting users during a holiday week or even selling loans can fit well in this type of communication.
We did user tests and were stunned by how well customers handled these situations: without any help, they were able to go through complex conversations, understand their options and make quick decisions along the way. The feedback was that they were more comfortable going through a banking product/service related topic this way then discussing it in person or over the phone with an agent.
Now we are moving forward, covering more and more topics and cases with this approach. Our short-term goal is to create a basic set of conversations to test with even more banking clients and I would say we are very much ahead on this road.
You want to see us sweating? Put our app to a test!
