Home Banking How the Banking Industry Can Capitalise on Conversational AI

How the Banking Industry Can Capitalise on Conversational AI

by internationalbanker

By Michael Conway, Data and Technology Transformation Leader, IBM Consulting UK & Ireland





Not a day has gone by this year, it seems, without a new headline about OpenAI’s ChatGPT or its equivalents. The world has been enthralled by stories about the novel ways these new tools are used, ranging from the impressive and imaginative to the downright bizarre. ChatGPT has helped make artificial intelligence (AI) real to many people who may have been unaware of its power.

Businesses, including financial institutions, have naturally viewed these developments with interest, wondering how to apply this technology to support their own digital strategies. There is good reason to be curious. The technologies underlying ChatGPT—generative AI and large language models (LLMs)—present a huge potential for banks to take their digital customer services and overall digital-transformation efforts to a new level.

It’s no secret that financial institutions of all kinds are racing to modernise digitally as quickly as possible to keep pace with their customers’ expectations and continue thriving in a more digital world. A host of challenges—a high-inflation, high-cost environment, the need to meet sustainability targets and the battle for scarce skilled talent—bring added impetus. Many firms have recognised the powerful role AI and automation technologies can play in helping them address these challenges, from removing costs by automating processes to using data-driven insights to track and reduce carbon emissions.

When it comes to using conversational AI, banks have already made brisk progress, particularly since the first COVID lockdowns rendered virtual customer-service assistants essential. Intelligent virtual agents, such as NatWest’s (National Westminster Bank’s) Cora1 and TSB Bank’s Smart Agent2, were deployed quickly to help bank staff handle the surging need for digital customer services, boosting the client experience and satisfaction in the process. More recently, IBM Consulting worked with Virgin Money to launch its digital host, Redi, to provide 24/7 support for the bank’s credit-card customers.

Making hyper-personalised services a reality

But this is just the beginning. Advances in generative AI and LLMs are opening up powerful opportunities for banks to achieve the kind of next-level digital customer experiences that will be essential for success in an increasingly competitive marketplace. Generative AI is capable of not only creating more humanlike, intuitive customer interactions, but it can also be tailored to reflect a specific brand tone of voice. Crucially, this technology will enable banks to deliver the genuinely hyper-personalised customer services that consumers will soon expect as standard.

Imagine if your banking app could help you automatically sort out problems proactively before you even knew about them, find the right products to meet your needs and provide advice that saves you time and money—all with humanlike responsiveness. How much easier would it be if that assistant could instantly calculate the best way for you to reach a particular financial goal, which might have taken you weeks of research, and make arrangements to help you get there?

And it’s not just online services that could be transformed. Generative AI and LLMs also hold the potential to help banks improve the customer experience in physical branches by assisting with tasks such as language translation in real-time for those whose native language isn’t the one spoken locally during often lengthy and complex processes.

Succeeding with conversational AI for business

However, unlike the consumer use cases we’ve heard a lot about, businesses need their conversational-AI platforms to meet a number of requirements to capitalise on them successfully.

Here are four important principles leaders should be aware of:

  1. It must understand the language of the business and the processes that make it useful to consumers.

An important part of productive conversational AI is understanding, relaying and applying business processes. While a generative-AI tool may be trained on massive volumes of text-based data, it does not necessarily “understand” specific business workflows. A bank wanting to roll out a customer-service assistant should invest time in refining language models around relevant financial terminology to increase the accuracy of the AI system so that customers receive reliable, consistent information.

Enterprise conversational AI should also be fully integrated into business processes. That means it must have the breadth of understanding to meet the user at any entry point or need and the depth of integration to help the customer through the whole journey, end to end.

  1. Details and clarification are essential components.

When your business is relied upon to conduct important financial transactions, provide essential services or deliver accurate information at speed and scale, there isn’t room for these systems to work only some of the time or be inaccurate, misleading or biased. That’s why the “best guess” approach of generative-AI tools for consumers is not suitable for business use unless applied with very careful consideration and guide rails.

Enterprise-grade conversational AI must have the ability to ask clarifying questions, disambiguate and understand additional details, as well as ensure the customer understands what actions will happen next. For example, if you go to a virtual assistant and say, “I need to pay my bill”, the assistant must be able to clarify which bill and from what account and also check that the customer is happy at every point before any action is executed.

  1. It should be highly personalised for the user.

Strong conversational AI deployed within an enterprise will know and understand you, then contextualise and personalise. It should also be able to tailor tone, voice, personality and linguistic elements seamlessly to each organisation’s requirements. Consider the example of cancelling a credit card. If you are using your bank’s virtual assistant, you would expect a contextual response detailing the process of cancelling your credit card with your specific institution. You would also expect your bank to have relevant information regarding your credit cards and understand the specific process related to the card you want to cancel. You’d also want the assistant to help you complete the process. It’s even better if it anticipates your needs before you even ask, like checking that your address is up-to-date before sending you a new card.

  1. It must be explainable, ethics-aligned and non-biased.

With a multitude of regulations in place or in the pipeline, financial-services businesses must be prepared to report on and defend, from an audit perspective, how specific models and AI-driven technology create their outputs. Those outputs must also be free from bias. At IBM, we have embedded our Principles for Trust and Transparency3 into how we develop AI offerings and acquire data for building and training models that power solutions, including IBM’s Watson Assistant service. Organisations should own their proprietary data and insights; thus, IBM clients do not have to relinquish their rights to their data to use our services. From the beginning, we have used responsible methods of generating data to train and optimise AI models because it’s critical that AI is used ethically. We bring this approach of trustworthiness and robustness to all of our deliveries, regardless of the technical stack.

Seizing the opportunity

Banks must create outstanding digital experiences for customers or risk being left behind. To make the most of the enormous opportunities presented by advances in conversational AI, leaders must look for solutions designed for the needs of businesses in highly regulated industries. They should be able to integrate with existing systems, incorporate specific governance policies and controls for accessing information within the business, and be specific to the company’s contextual requirements.

Finally, working with partners that combine extensive banking-industry knowledge with deep technical and business-transformation expertise will offer financial institutions the best chance of success. Delivering delightful and novel experiences using cutting-edge AI techniques in highly regulated environments is not easy. With the right partner, however, the outcomes are incredible, and the time to do it is now.



1 IBM: “At Your Service: ‘Cora’ Helps NatWest Deliver Exceptional Customer Support,” Jake Collins, Tim Quigly and Tony Hickman, June 27, 2022.

2 IBM: “TSB Smart Agent goes mobile with IBM Watson,” March 25, 2021.

3 IBM: “IBM’s Principles for Trust and Transparency.”


Michael Conway is an Executive Partner of IBM Consulting responsible for the Data and Technology Transformation Service Line in the UK and Ireland. He is accountable for the Service Line’s strategy, execution and growth, focusing on ensuring that IBMers can build great careers whilst delivering great outcomes for clients.


Related Articles

Leave a Comment

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.