Bots can manage thousands of intentions nowadays, which might already seem like a lot but in fact it’s far from what specialists believe it can achieve. However, what ensures a good customer service is not just the number of intentions it can manage but also knowing how to separate what the client's intention is so that there are no misunderstandings. Additionally, as society advances and new intentions are added, it will need to learn to incorporate new intentions and differentiate them, avoiding mistakes and ultimately, becoming a large and uncontrollable artificial intelligence bot.
While many industries are more inclined to implement such disruptive technologies, the digitization of the banking sector is moving at a different pace because of security reasons and having to deal with very sensitive information. However, as customer interaction between brands and clients is migrating from offline to online because of the growing popularity of social media platforms and other conversational channels such as WhatsApp, the banking sector has had to adapt.
What is the new “bot-of-bots” model and what are its advantages?
In their journey to digital transformation, banking institutions are relying more and more on virtual assistants to solve customer needs in an agile way. The rapid responses that AI bots can provide and how they relieve human representatives of certain repetitive tasks, are some of the most important advantages of virtual assistants.
Nonetheless, the lack of a high level of maturity of these technologies has proven that, in several cases, a virtual assistant isn't enough to find the right solution for a problem presented by a client. Consequently, a new interaction model between bots has been emerging to increase the efficiency of virtual assistants. The system relies on a parent-bot, or megabot, that identifies the client’s demand and delegates the task to another bot that has been trained in that specific area. As a result, the response to the issue ends up being much more precise and thus, efficient.
Adopting a bot-of-bots model is an upgrade in customer relationship management because of several benefits. Due to the fact that a query is solved in a shorter period of time, allows for more claims to be dealt. The compartmentalization of the bots according to their area of specialization helps avoid the saturation of a singular virtual assistant. Also, a bot-of-bots system makes it possible for every particular business area to take control over their specific bot when needed.
The challenges behind the innovation
Given the technological complexity of a bot-of-bots model, adopting such a system in a bank presents many challenges.
First and foremost, the implementation of AI bots needs to be carefully studied because of the sensitive data that is being handled. A simple misunderstanding could lead to a significant problem for both the client and the bank.
More frequently than not, the collision of the concepts involved in conversational banking can happen, and if not given enough attention, different intentions could end up being activated with the same text. Asking a bot for help to complete a check-in isn’t the same as asking where the check-in can be done. In this context, having a reliable cybersecurity system in place will be key to reducing the number of problems and their potential impact.
Bots system training
Secondly, these technological systems aren’t solid enough to implement just one cognitive training model. For this technology to be a success, multiple cognitive engines should be adopted in order to make good use of whichever is more adequate for every particular occasion.
The focus of convergence should be put on the routing systems that are in charge of connecting the multiple bots, while delegating the issue addressed to the proper child-bot. The correct development of searching tools that allow the megabot to identify the root cause of the issue is what can ultimately help provide a better and more efficient customer service. A data analytics system that can collect and synthesize all the information can be very helpful in defining this process.
Humanization of bots
Additionally, it is vital that the tone and behavior of the virtual assistant remain constant in order to make this model work. A natural state of conversation between the bank and its clients can only occur if the bot sounds coherent and natural.
For this to happen, the combination of information gathered by human representatives along with this new automatized system is vital. Social media management and face-to-face interactions with the clientele are invaluable inputs for the creation of a more accurate bot training model.
During the podcast dedicated to why the bot-of-bots technology is the future of interaction between a bank and its customers, Fabio Distaso, Head of Conversational Banking at NTT DATA EMEAL, Beatriz Albert, Conversational Banking Leader at NTT DATA EMEAL, and Mario Armas Ramírez, specialist in the world of conversational assistants, have reached a few key conclusions:
- The new interaction model between bots relies on a parent-bot, or megabot, that identifies a query and immediately delegates the task to another, more specialized, bot that has been trained in that specific area.
- Among the main benefits of a bot-of-bots interaction model is increasing efficiency and therefore, customer satisfaction, compartmentalizing bots according to their area of specialization which helps avoid the saturation of a singular virtual assistant, and making it possible for every business area to take ownership of their specific bot when needed.
- Compliance and security concerns related to the sensitive nature of the data that is being handled, the technological systems that aren’t solid enough to implement just one cognitive training model or managing to set a constant tone and behavioural pattern for the virtual assistant, are among the main challenges of implementing a bot-of-bots interaction model.
All in all, bot-of-bots technology can be a major evolution in the digitization of the banking sector and, as the various challenges are successfully resolved, the speed of implementation of the new models will bring major improvements at various scales.