Chatbots and Banking 101: a lightning-fast journey to conversational banking
As we increasingly rely on internet connected devices to manage our lives, we are naturally exposed to interactions with ‘conversational user interfaces’ or chatbots. In serious matters like banking and finance we would have expected a natural distrustful penchant for bots, yet most of the banks and their clients are undoubtedly not foreign to robotics. This is because bots can provide a great deal of service improvements to both clients and institutions.
For once, customer service has usually been far from optimal. Bots can radically improve the status quo in many ways. They respond faster, do not experience peak times or long days, can provide assistance on a 24/7 basis, can improve fraud prevention, and can collect vast amounts of useful information about you. These data collected are crunched to anticipate your future needs and deliver a more personalized experience on each and every banking interaction.
Personal banking services most bots currently provide are limited to basic information requests or transaction related queries. Examples are a client asking for the next monthly amount due of a loan, details of a payment, the total amount charged via direct debits last month, or the number of outstanding direct debits. Assuming all sort of services could be delivered, bots would become the single digital point of contact for banking matters.
One can imagine these interactions evolving towards a full human to human experience in that bots could be requested to execute payments, set up direct debits, provide leads on market products and services, look for the best market deals or, by gleaning relevant information from spending patterns, to assist clients in managing their finances.
From a bank point of view, more client facing interactions being handled by bots allow operational resources to be reallocated to alternative value-added activities, with bots and staff occasionally assisting each other as the situation requires. Let’s imagine a back-officer relying on bots generated answers to respond to clients’ queries, or a bot providing an irrelevant answer and the back-officer acting as a smart interface between client and bot.
Further to client service, bots’ capabilities can also be leveraged to help employees with clerical tasks, required training, or IT and HR requests among others. In the same vein, bots can host educational videos to assist clients with basic banking tasks, inform them of banking updates, or being used as a marketing channel.
Burdening-off client service desks reduces client service costs while more clients are assisted quicker in a more personalized manner. In theory, these service improvements would also translate into further client engagement, a strengthening of long-term relationships, and could indirectly attract new clients to the bank.
The most interesting benefits however, do not belong in the cost equation but in the revenues one. As a wealth of clients’ data are collected, behaviours exhibited, questions asked, and preferences shown can be analyzed. Clients’ demands, satisfaction levels, personal interests, etc. are therefore measured more accurately and understood deeper. This paves the way to a fully client-driven design of personalized products and services delivered at the right time and in the right fashion.
In the insatiable quest for service improvements, the cycle of capturing and analyzing customer engagement data to deliver tailor design products and services would continuously repeat itself.
Banking chatbots today
Banking bots today do not fully meet clients’ expectations but are fast reaching satisfaction levels. Clients’ frequent complaints highlight difficulties to making themselves understood and to maintain a conversation with bots especially when slang is used, ‘unexpected’ questions are asked, or several questions are asked at once. Furthermore, voice-enabled bots have a hard time trying to deal with for example, accents and / or moods. This is of course not exclusive to banking bots. See this funny video featuring Amazon’s Alexa below:
Natural language learning is crucial to improve the contextual appraisal of conversations and transform the current client experience. This learning fits bots with the capabilities to address non-structured, complex questions, and improve the quality of conversations. It is in great part thanks to the advent of Natural Language Processors (NLP), as a service platforms, that many chatbots have been successfully introduced into the market.
Current language processing capabilities however, cannot yet guarantee autonomous responses when it comes to addressing complex questions or resolve disputes, with clients preferring to talk to humans rather than bots in those instances.
See in Fig.1 below where a Natural Language Processor sits within a high-level conversational architecture interfacing banking users and back-end banking applications. The Natural Language Processor plays the role of translator whereas the intent engine, supported by Artificial Intelligence capabilities, does interpret users’ queries and ‘intentions’.
The way Artificial Intelligence helps humanizing banking is by joining efforts with NLP through aggregation, augmentation, and summarization:
- Aggregation: aggregates data from every conversation and draws insights, identify trends, and more importantly, learns to predict and anticipate clients’ needs
- Augmentation: augments human capabilities with bots stepping in to assist humans in a collaborative model as above mentioned
- Summarization: automates summaries creation while obtaining same results as if those summaries were written by humans
Boosted by AI, the intent engine thus retrieves the relevant information from the banking back-end applications and respond back to clients. Finally the conversational persistence data base in the figure, historicizes conversations usually relying on a cloud system to limit data storage issues.
Banking chatbots tomorrow
Whether these are text or voice-enabled, there is clearly no steep learning curve to using conversational interfaces. Bots are easily and discretely embedded into proprietary online channels or more conveniently, in social media platforms such as WeChat or Facebook.
Of course financial data are extremely sensible and potential security issues are of the utmost concern. Though bots require limited support, it is extremely important they undergo extended security and performance testing to prevent harmful breaches.
In almost every respect however, bots fare much better than commonly used banking channels. While a bit of learning to understand the user interface is required to interact with traditional channels, bots learn from you instead. Conversational banking is naturally easier, does not require downloading or installing any piece of software, and provides the best possible personal advice eventually surpassing humans.
Bots’ learning speed will be the most critical factor to ensure their future success. This learning is fast progressing in two separate but complementary streams: the first one is aiming to enhance current capabilities in clients’ analytics, AI’s aggregation function, and the second is focused on improving understanding of natural language. Both combined, can work effectively towards delivering client-driven personalized banking products and services.
As banking clients we are witnessing the evolution of bots capabilities thanks to this continuous learning. Not so long ago, bots were barely able to respond to simple queries whereas they are now able to perform banking tasks and managing transactions. It is envisageable that conversational bots will shortly be able to feature further advising capabilities, thus fully humanizing the banking experience and shifting the business from the current transaction based model to a conversational one.
The future holds great promise with bots leading us into a world of voice authentication and facial recognition capabilities. They would likely feature virtual reality simulations and will be integrated with IoT capable devices and blockchain platforms. Though venture capital funding for AI startups is at record highs, a faster development of bots is being hindered by a short supply of AI skills still catching-up with demand.
Some years from now, will we be able to tell we are banking with a bot or a human and, what are the implications for both clients and institutions?
