Upgrading Virtual Assistants with Machine Learning | NTT DATA

Thu, 29 September 2022

Upgrading Virtual Assistants with Machine Learning

Learn how machine learning is allowing the banking industry to offer humanized and hyper-personalized virtual assistance to customers.

Innovation, security and ethics: How machine learning is taking banking virtual assistants to the next level

Humanizing banking virtual assistants through machine learning

Did you know that most digital translating tools rely on machine learning? In fact, this branch of artificial intelligence is leveraged in many popular applications such as estimating a commute, filtering spam emails or training virtual assistants. Deep-learning models, which are also called ‘transformers’, are already capable of generating texts and data that seem coherent at first glance but lack accuracy. These learning models are becoming more and more important in the development of automated tools that can help companies reach their customers digitally, such as virtual assistants. However, the uncertainty of their accuracy is not acceptable in industries that deal with delicate information such as banking institutions.

The new relationship models that are being established, thanks to machine learning, between bank bots and their clients, are posing new challenges that need addressing. Virtual assistants must be able to deliver correct answers about specific data because any error might have serious consequences on the consumer’s finances. This can cost an institution the trust of its customers, many of whom are reluctant to use technology to begin with. In order to regain the feeling of safety around sensitive topics such as personal finances, machine learning should be strictly focused on humanizing the conversational processes. Therefore, turning conversational banking into a trustworthy, secure and humanized mechanism has become a priority when it comes to transformers.

Will the banking industry be able to successfully move past these challenges and fully capitalize on these learning mechanisms in order to upgrade their conversation with their customers?

How can machine learning improve virtual assistants

As mentioned, when it comes to conversational banking, the greatest challenge is the diversity of the clients and the way they communicate. By implementing advanced technologies such as machine learning in the customer service strategy, banks can offer hyper-personalized services for more specific needs. Data gathering, the algorithm analysis and the patterns obtained through AI, can enhance the quality of the answers that are provided by a bot to a client.

Additionally, these technologies aim to achieve a fully automated communication strategy that correctly perceives the client’s context and mood while also understanding the requests in a more precise manner. Therefore, machine learning adds strong value to the conversation by creating more humanized responses. The capacity of adapting to every client’s particular reality is what positions machine learning as the ultimate tool for creating organically better, more authentic interactions between the bank and their audiences.

New challenges for the banking industry: building a secure and ethical system

Although machine learning is proving to enhance the conversational bank experience, there are still some issues that need to be adjusted in order for it to function perfectly. Having to communicate with people of different ages and backgrounds, and provide a personalized service that caters to such a wide public is a very difficult task.

Given the delicate nature of the conversations, where clients might inquire about payment terms or credit information, the challenge isn’t only in communicating effectively with a diverse audience, it’s also in making sure that the information is correct and shared in a safe space. In fact, the delivery of a precise result is often vital in certain machine learning applications. When it comes to bill payment deadlines, for example, a mistaken piece of information could lead to extra charges or penalties. In order to make customers trust virtual assistants, the level of accuracy that is required in the correct implementation of transformers needs to be as high as possible.

Transformers - what they are and why they’re popular

A transformer model is a type of neural network architecture that is becoming more and more popular due to its incredible results in comprehension, text generation and other high volume processing tasks.

Although it is estimated that a single training session provided by transformers can cost up to 12 million dollars, many organizations and specialists prefer them to other solutions. However, they come with some clear limitations. Due to the massive volumes of information they process, transformers can easily end up being contaminated by different biases. In order to process curated information in an ethical way, bots need to learn beyond a simple code sequence. Hybrid deep learning models are a viable solution because they merge formal content learning with contextual information that encompasses the correct use of the formal concepts.

Concerns about the ethical use of advanced technologies have been raised by specialists for quite a few years now. As a consequence, in April 2021 the European Union implemented a very strict series of norms for the ethical use of AI in order to properly differentiate a simple algorithm from a whole machine learning system that can ultimately be in charge of deciding, for example, if a loan is granted or not. These restrictions are meant to challenge the use of AI and analyse the systemic structures that work behind these great technological devices in order to implement them fairly among societies.

During the podcast dedicated to how machine learning is enhancing virtual assistants, our guests Beatriz Albert, Solutions Specific Knowledge Analyst, Digital & Retail Banking and Fiorela Doti, Analytics Specialist together with Fabio Distaso, Head of Italy & Global Head of Conversational Banking at NTT DATA reached a few key conclusions:

  • Specialists believe that blockchain-based data distribution combined with the hybrid learning models will eventually lead to a more democratized implementation of artificial intelligence around the world.
  • The future of machine learning will rely on the interoperability of evolved transformers created by specialized professionals such as language data scientists, computer linguistics and blockchain professionals.
  • In order to help the banking industry achieve the perfect customer experience while also providing secure, ethical and humanized services, it is vital to make sure that the correct training program is being implemented.

The conversational banking experience will be close enough to achieving an ideal level of empathy, understanding and humanization, only when solid background information, propper education around the topic, and the help of specialized professionals are correctly applied.


How can we help you

Get in touch