GenAI or replatforming? Bank CTOs disagree on budgetary spend

From Paul Julius Reuter’s embrace of telegraph lines to receive financial information, to Edward Calahan’s creation of the ticker tape in the 1860s, technological disruption has been a constant throughout the history of the capital markets. And while it might have seemed at the time that these advancements came out of nowhere, they were actually the result of technological evolution. 

Fast-forward to the start of 2023, when the ChatGPT chatbot started appearing more frequently in the headlines of technology publications and scientific papers (and way too many LinkedIn posts). By the end of the year, generative AI and large language models were on the lips of not just technologists, but bank, asset management and exchange CEOs on earnings calls. But genAI and LLMs are not new – they’re merely the culmination of decades of evolution in the field of artificial intelligence combined with numerous other disciplines.

Slowly, financial institutions are turning into tech and data companies. At the same time, the pace of change is increasing so fast that it’s difficult for even the most plugged-in CTOs to stay abreast of these tectonic shifts, making budgeting exponentially trickier. Take, for example, ChatGPT: As bank CTOs were planning their budgets at the end of 2022 for the coming year, genAI was unlikely to have been a hot topic. What a difference a year makes.

When you have the AI story to tell, it becomes a much more compelling story, but that can’t help our real goal, which is cloud

Chief information officer at a G-Sib bank

“I would have to say that for the year of 2023, AI was almost exclusive to everything we did,” the head of engineering and technology at a global systemically important bank (G-Sib) tells WatersTechnology. “Of course, we’re delivering on regulatory obligations and long-term projects, but in tandem with that, we’ve put an uber amount of focus on leveraging that data and taking in the advanced side of AI – i.e., generative, large language models and cognitive compute aspects. We have accelerated that significantly in 2023, and it’s going to be a big story for 2024.”

What many banks are learning, however, is that while they may want to embrace cutting-edge tools, beyond regulatory and risk concerns, there are significant barriers that need to be addressed before AI nirvana can be achieved. 

Push/Pull

“We have been looking at blockchain, but I think that’s a solution looking for a problem,” says the capital markets CTO at a second G-Sib. “Since 2016, for most use cases, it’s been a complete waste of time.”

This insight illustrates the challenge CTOs face when allocating dollars to not falling behind on the innovation front, while not wasting resources that could have gone toward modernization, automation, data management and regulatory needs. But when it comes to genAI, the bank CTO is all in. 

“There’s no one technology that we’re aggressively pursuing because it’s solving so many problems for us,” they say. “But when it comes to generative AI, we’ve started spending a lot of time thinking about how to use that.”

Still, the chief information officer for capital markets at a third G-Sib takes a more pessimistic view of genAI – or rather, they believe that exploring genAI will sap resources from the bank’s cloud efforts. The bank still relies on a variety of legacy systems that are accruing technical debt and hindering data aggregation efforts. 

While AI is a useful tool, it relies on a wealth of data, especially when it comes to genAI or LLMs, versus something like robotic process automation (RPA). The more applications that can be re-platformed to the cloud, the more flexibility the bank has to build microservices rather than technical-debt-laden, monolithic platforms.

When done well, re-platforming on-premises systems and platforms to the cloud provides for increased scalability, improved costs related to technical debt and the ability to develop microservices, among other advantages. On the other hand, data migration issues are disruptive, integration with other systems can compound complexity, and while (in theory) cloud is cheaper than an on-premises monolith, managing cloud costs can be tricky.

“The problem is when the time comes and they want to migrate 50 applications and the business says, ‘Why would I do that versus addressing this big regulatory obligation I have?’ Or, ‘Why would I stop some revenue-expanding functionality to focus on the cloud?’” says the CIO. “Well, when you have the AI story to tell, it becomes a much more compelling story, but that can’t help our real goal, which is cloud.” It’s similar, they say, to how blockchain opened up purse strings for under-invested back-office needs in the mid-2010s.

A fourth bank CTO similarly laments that the business side of the organization rarely understands the impact of technical debt on being able to invest in more forward-looking initiatives. But the business side is open to cutting-edge new technologies – which can increase alpha but which they know little about – versus the very boring topic of technical debt. 

Technical debt has always been an issue for technologists, but following the global regulatory overhaul brought on by the 2008 financial crisis, banks – which had until then been more likely to build than buy – had to turn to the vendor community to help with a raft of new reporting requirements. The problem was that the banks were encumbered by decades-old legacy systems and platforms that weren’t interconnected. They started to realize the depth of the hole they had dug for themselves. 

The goal today, says the fourth CTO, is to get to the cloud because if you can break down data silos and therefore connect systems in the front, middle and back offices, you have data that is more consistent, of a higher quality, and that can be used for multiple purposes – whether that’s for trade analytics, risk management, or regulatory reporting needs.

“There are many people far better and wiser than me, but we have tried [getting budget] in terms of getting those data lakes/warehouse and data enterprise budgets – data governance, data quality – those big pieces whereby we need to clean out the backend because, you know, shit data in, shit data out. Makes perfect sense, right?” the fourth CTO says. “But I can’t think of many that have been successful … because the business doesn’t buy into it.”

Back in the 1860s, some financial institutions scoffed at telegraph lines and ticker tapes. Eventually, those inventions revolutionized the capital markets. They also laid the groundwork for computerized trading and the New York Stock Exchange’s Designated Order Turnaround, which came about a century later. That led to the rise of electronic communication networks and high frequency trading and, today, the world of digital assets and cryptocurrencies. 

Technological evolution has been a constant since the inception of the capital markets, but the pace is accelerating, making it more difficult for banks to keep up. That means that now more than ever, a CTO’s job is to balance day-to-day and regulatory needs against the risk of the bank falling behind in the race to capitalize on new technologies.