With the use of mobile-messaging apps now outstripping the active use of social networks, forward-looking financial service providers (FSPs) are rethinking the way they interact with customers. As bots can be integrated with messaging platforms, FSPs are increasingly considering the deployment of chatbots to augment customer service.
Chatbots have been described as “software that can have a conversation with a human” (but not necessarily in the way that a human would converse with another human) and among other things, they allow for a virtual agent to be available to customers 24/7.
Robo-advisors are distinct chatbots that (within the financial domain) offer regulated financial advice on specific topics (e.g. investment advice).
Companies (from within and outside of the financial sector) are turning to chatbots for numerous reasons, and the use cases extend beyond narrow customer service applications. Some of these include:
- Accessibility (bridging the gap between tech and touch). Chatbots assist in providing convenience to the customer (the 24/7 availability mentioned earlier). Furthermore, mobile-phone users, particularly younger users, increasingly prefer texting to making voice calls, and a chatbot that has been integrated within a messaging app is both immediately accessible and convenient.
- Inherent advantages over interactive voice response. Due to advances in natural language processing, users can interact with chatbots in a way that imitates more “normal conversation” patterns. In other words, they can more closely mirror an in-store or in-bank experience. Enhanced flexibility means that customers aren’t necessarily confined to a menu of predefined options and answers.
- Efficiency and cost-saving. Chatbots have (with defined limitations) the potential to outperform people in terms of speed and accuracy of information provided. Bots can be very effective in dealing with more routine questions that are frequently asked, freeing up the time of call centre agents so that they can devote time to answering the tougher questions or handling elevated complaints.
- Expanded financial inclusion. Accion outlines several ways in which chatbots can contribute to financial inclusion; including enabling individuals to improve their own financial capability.
- Improved response rates. Chatbots can be effectively deployed as email substitutes, and open and clickthrough rates outperform traditional marketing emails. This is one outcome cited in an article on the success of “Emilia” in Columbia (a chatbot created through a partnership between Bancolombia and Juntos).
- Data collection from the field. Read the i2i report on spatial data for context on how “Bots can be programmed to function in multiple languages and can collect a range of data types, including location data, pictures, video and audio recordings.”
The buts of bots
Amid the hype about bots, it’s easy to forget that the effectiveness of bots still has limitations. These limitations or challenges include:
- The relative immaturity of the underlying machine learning and data analysis. Chatbots are relatively new in terms of their use within the financial sector, particularly the financial sector in Africa. Absa (part of the Barclays Africa group) launched Africa’s first banking bot in mid-2016.
A comment in this blog from Zendesk offers a sobering perspective: “… the data needed to train those bots to reply in a way that’s even vaguely human is still a few years out of reach for most. The titanic three – Google, Facebook and Apple – have access to most of that data, and their army of engineers still haven’t managed to completely close that gap”. |
Before developing and deploying a bot within your financial service, you would need to ensure that your underlying technology and data practices are sophisticated enough to support the algorithms that underpin it.
- Uneven experiences are still the norm. In this interesting interview, Peter Wannemacher (referencing a now two-year-old Forrester report) says that bots are only effective and successful 70% of the time. As he explains, “The problem is that 30% [the portion of the time that they are ineffective] is really high for banks and add to that the fact that people’s money is an especially sensitive subject.”
- Humans are still needed as part of the customer service offering. In the Zendesk blog referenced above, the author recommends that bots be programmed with an “escape hatch”. Essentially chatbots are not able to completely plug the customer service gap, and a smooth handover to a staff member who is capable of extending the service or dealing with an atypical issue should be a carefully considered part of the offering.
Before you bot
The articles below reference some of the key considerations for FSPs that are planning to develop and deploy a chatbot:
- If financial inclusion is your aim. Accion has listed some attributes that they look for in financial chatbots – the first of which is “narrow application”. They cite the example of the insurance industry, where the successful use of chatbots can lower the brokerage costs of providing services to lower income groups.
- Your data needs to be “trained”. A prerequisite for successful chatbots is training one’s chatbot so that it gets “smarter”. This includes training for semantics, personalisation and accuracy. Read more on data training here.
- Security features should be incorporated. In a blog from Abe.ai, reference is made to the fact that although chatbots still operate on standard internet protocols, secure authentication and authorisation procedures should be carefully considered in the development process.
How to bot
If you’ve decided to go ahead and develop a bot (rather than partner with a commercial bot developer), here are some practical resources that can serve as a starting point:
- A chatbot blueprint. This Deloitte article explains the differences between informational, transactional and advisory bots and includes five personas for which to design a bot.
- Basics for building a chatbot. Marsha Hunt writes about the attributes of a successful chatbot – including giving your bot some personality.
- Four requirements for custom bot development. Read more about the four pieces of tech infrastructure needed to custom-build your bot. Essential considerations are the messaging interface for users, the natural language processing engine, a chat interface to respond to escalated messages and the bridging software to connect these.
- Best chatbot-building platforms. For more technical insights, here is a handy list of platforms that chatbot developers can use.
Please note that sharing links to publications or articles that are not authored by i2i does not constitute official endorsement of or agreement with the content contained in those links.