Banks, genAI & the big picture

They say a picture is worth 1,000 words - but in some cases, it can feel like even more than that.

There was a widely circulated chart on social media in 2023 that sketched out the speed of adoption of ChatGPT since the generative artificial intelligence program (genAI) had launched. The chart – which you can see here – visualises the rapid pace at which ChatGPT hit a userbase of 100 million people (just 60 days in this telling); significantly  faster than hugely popular online tools like Instagram, Spotfiy, and Facebook. This picture showed ChatGPT was something new – and something big.

For many businesses, particularly banks, it’s a picture that could be worth much, much more than words. And it’s on that basis entire industries around the world are spending time – and money – looking into how they can leverage generative artificial intelligence technology to take the next step into the digital era.

 So how much, really, could it be worth?  Well, a 2023 study from McKinsey suggested productivity increases from the use of genAI could lift annual industry revenue by between 2.8 per cent and 4.7 per cent – or $US200 billion to $A340 billion. That’s more than 1,000 words.

The questions for banks, then, are simple: how can they take advantage of this evolution in technology, and what are some of the use cases that could apply to their businesses? The answers all start with data.

Engine

It’s well known data are what fuels AI, whether that’s traditional machine learning or more modern genAI models. This gives banks a natural head start over many other sectors, given how much data they produce and securely store. For the financial services sector at large, the opportunity could be significant.

In the same vein, taking advantage of that opportunity could bring challenges - and those would need to be addressed before the benefits of genAI technology are seen.

It’s a common refrain to hear banking described as information rich but organisation poor. While large volumes of data are generated in the sector, most of them exist on disparate systems that don’t talk to each other, and are sometimes inconsistently applied.

Additionally, within select banks there are various degrees of data maturity and digitisation, which can often reflect digital journey of that bank’s customer, in a lot of cases. All this makes coming up with comprehensive solutions a very complex task.

That complexity, however, is the exact kind of problem genAI is designed to help address.  Models trained for a variety of tasks, rather than traditional models focused on specific tasks, are capable of handling large volumes of unstructured data across formats (such as documents, images, pdfs, and the like) and provide output in a simple, conversational format for anyone to use.

While the quality of data still determines the quality of output, the technology provides the opportunity to explore the art of the possible – allowing accurate information retrieval at the time you need it.

Generate, rather than just predict

Banks have been using AI for a considerable length of time. Traditional machine learning (ML) models have been used in a variety of use-cases ranging from meeting know-your-customer requirements, transaction validation and fraud detection, to other things like assessment of credit worthiness. But genAI is inherently different.

For staff at a financial services group, genAI can draw on large volumes of existing documents to deliver specific information from the latest policy documentation that considers multiple factors.

Previously such an activity could only be accomplished through human centres of excellence, which increases the risk of key person dependencies. There’s a large opportunity for genAI to address this issue.  

Enhanced customer experience and creation of customised content at scale is another arrow in genAI’s quiver. For banks, the use of genAI could elevate the customer experience through tailored and context-aligned communication, due to the ability of large language models, (LLMs) to assimilate information, make connections and inferences.

The technology could help banks to respond to customer queries and request for proposal documents, due to its ability to aggregate and synthesise information. This may drive more efficient, accurate and customised responses, paving way for stronger relationships between bankers and their customers.

In some cases, given the multi-modal ability of genAI, the technology could provide for the generation of hyper-customised marketing collateral – including text, images and video.

Speed of technology delivery is another area genAI could help banks. One of the functions in which genAI has shown consistently great results is its ability to generate code.

Modern banks are predominantly technology led and have been migrating away from old legacy technology systems all around the world. The ability to interact with LLMs using natural conversational language to generate code has already proven to be one of the biggest advantages of genAI. This use case could see a rapid uptake to speed up delivery.

And these use cases above are just the tip of the iceberg when considering what is possible though genAI.

Execution

Of course, the key challenge to deriving value from any new technology is less about designing use-cases, and more about how you deliver on them.

It is widely believed genAI is a technology that can deliver meaningful, customised and ‘moment-in-time’ customer experiences, but success will rely on organisations taking a holistic, business-led, customer-centric approach to the technology, and not treating it just as another in a series of tools.

This means the technology would need to become a strategic lever within a business, allowing operations and technology teams to work together toward a common goal.

For banks considering use of the technology, there are some key factors to consider before adopting the tech that would help get the most out of the opportunity.

The first is a clear vision for how the technology could help; a ‘north star’ that will bring the whole organisation together and align the tech with strategy across all levels. A clear, time-bound roadmap with well-defined milestones to drive the strategy forward is also critical here.

Banks should ensure they have commitment from the staff on the use of the technology and everyone has ‘skin in the game’ on its success. A clear communication plan for taking employees on this journey is important.

When it comes to execution, an open innovation model is critical. Putting the technology in the hands of a company’s people, helping them learn to use it safely and effectively, and allowing them to innovate, is the best path forward, rather than taking a top-down approach.

As with any new technology, regulatory engagement is crucial. And this goes hand in hand with a strong risk and compliance framework embedded every step of the way, along with very strong controls and governance.

Given banks operate in a stringent regulatory environment, the path to success is not about just being first, but also about being right.

Madhujith Venkatakrishna is Associate Director, Innovation at ANZ Institutional