Icy attitudes on internal GenAI usage are thawing—and just in time

Waters Wrap: More and more banks and asset managers are publicly talking about how they are experimenting with generative AI. In the fight for talent, Anthony says this is a necessary step.

Credit: Charles Frederick Ulrich

“The discussion in public has been pretty much AI being the job-killer number one—it’s definitely not. It’s rather leading to a new pattern, where you actually need more skilled people because they not only have to understand their current and existing models the classical way, but they also need to understand what the AI does with them. And only that combined knowledge actually allows them to judge whether the output is correct or not.”

These are the words of Wolfgang Koehler, chief risk and compliance office at Mizuho Securities Europe. He was speaking at an event in Germany run by WatersTechnology’s sibling publication, Risk.net. The conversation up on stage was about how financial firms are warming up to the use of large language models and generative AI.

In July 2023, we published a story about how at least six global systemically important banks (G-Sibs)—JP Morgan, Bank of America, Citi, Goldman Sachs, Wells Fargo and Deutsche Bank—had temporarily curbed the use of ChatGPT among employees. Morgan Stanley was, at the time, allowing for limited personnel access, but only for pre-approved use cases under rigorous governance and controls.

There are four major concerns when it comes to generative AI: data security; copyright issues; explainability; and “hallucinations.” Financial institutions are heavily regulated, so banks were naturally taking a slow, considered approach to rolling out these powerful-but-young tools.

“We are still at an exploratory stage to ensure the technology aligns with our risk appetite and regulatory requirements. We want to use it in a way that we can explain the technology internally to ourselves and create real economic values for our clients,” said the head of AI at one of those banks.

But there are signs that these somewhat chilly attitudes towards GenAI are starting to thaw.

At the Risk.net conference, Koehler noted that Mizuho’s parent entity had “already rolled out AI to a good 40,000 employees in Japan,” but is still figuring out how to roll out GenAI in the US and Europe. And that’s something that every bank will need to figure out if they want to stay competitive in attracting talent.

“I think being on that journey is absolutely critically important because whether you look at someone in an operations area or someone in risk management or the front office, every entity wants to utilize AI. They are using AI already in their private lives. … We need to get there because, ultimately, it will be a huge differentiator when you’re trying to attract staff. When you tell them, ‘No, AI is forbidden here,’ that’s certainly not the answer you want to give,” Koehler said.

While delivering the opening keynote address in September at the World Financial Information Summit, Marion Leslie, head of financial information at Six Group, echoed a similar sentiment: “Now is the time to define your path through the too much/too little paradox, because the one thing you don’t want to do is be too late.”

Vibhor Rastogi, global head of AI, machine learning, and data investments at Citi Ventures, Citi’s investment arm, told WatersTechnology in August that investing in the technology has become table stakes for financial institutions, but he says that the firm has been “very thoughtful” about its use of AI. “I personally think that every financial institution has to invest in AI and use AI as a business imperative. Just like prior technologies like cloud, mobile, and others, I think financial institutions are seeing that they could be competitively disadvantaged if they don’t use AI, so I think in that sense, yes, it’s becoming table stakes.”

In May, Nitin Tandon, chief information officer at Vanguard, told WatersTechnology that the asset manager was experimenting with GenAI with the goal of increasing the efficiency and productivity of the company’s advisers.

Last year, Vanguard rolled out GitHub Copilot to its developers, allowing them to write code more effectively, and GenAI platform Writer is being used for content creation in marketing and blog posts on the company website. “There’s so much potential in making our advisers more productive,” Tandon explained. “A financial adviser spends a lot of time preparing for calls, and then afterward has to write down key takeaways. We just saw the results from a model we developed that summarizes these calls, and that’s fantastic to see. This so-called summarization is something we’ve been working on for our advisers.”

Then in April of this year, Sathish Muthukrishnan, chief information, data and digital officer at Ally Bank, explained how Ally decided at first to bar use of ChatGPT throughout the bank, with the exception of a lab function within its technology group, where experiments could continue. “That was not a popular decision, but it was the right one,” Muthukrishnan conceded.

Muthukrishnan worked hand-in-hand with Ally’s chief risk officer Jason Schugel, who decided that rather than treating GenAI as a model risk problem, the bank should approach it as a product risk problem, funneling each instance of GenAI through Ally’s existing new product approval process.

This had a number of advantages, chief among them the presence of business lines and corporate functions at various stages of the review process, as well as second-line risk managers who collectively cover the 12 forms of risk—and 30 “child” exposures—that Ally identified as its material risks. The block was removed, no new bureaucratic layers were created, and uses of GenAI were suddenly exposed to scrutiny from a whole range of experts and stakeholders, rather than being the preserve of model risk specialists.

“It was a stroke of genius,” Muthukrishnan said. “The new product committee has constituents thinking about every risk dimension across the company—12 at the top and then multiple sub-levels. It was a very simple idea, but it was also very powerful. Everybody that is part of the committee understands the risks they are assessing. They understand the process through which they assess the risk. They know the questions to ask. And that was what allowed us to accelerate from experiment to execution.”

Also in April, Man Group CTO Gary Collier said that he foresees generative AI “underpinning” entire workflows at the firm, starting with nascent areas such as time-series analysis in the quantum systematic space, before expanding to cover other areas, complementing its existing AI copilot, Alpha Assistant.

“You could take the view that it’s not really applicable to the quantum systematic space, but if you take a step back and look across the entire value chain of everything from hypothesis generation, dealing with data, and easy interfacing into code and into data, it’s a binding agent,” Collier said. “There have been use cases of generating synthetic time series. I see [GenAI] very much underpinning the entire process.”

In fact, one of the things that differentiates AI from other technology projects is that people across all those workflows and business lines want to be involved in its development.

“One of the interesting things about GenAI is that for the first time, something that is at the leading edge of tech is not wholly in the domain of quants. There’s loads of interest in use cases across discretionary and other parts of the firm as well, which you may not have thought of as wanting to be at the cutting-edge of tech,” Collier added. “So it very much is, I think, more than anything we’ve seen in at least recent history, an enabler. And that’s novel.”

Let me go back to what Mizuho’s Koehler said previously: “We need to get there because, ultimately, it will be a huge differentiator when you’re trying to attract staff. When you tell them, ‘No, AI is forbidden here,’ that’s certainly not the answer you want to give.”

This reminded me of an article we published last week about Bloomberg making a significant investment to hire quants with PhD and MSc qualifications to improve its quant pricing and research products. Max Bowie reported that Bloomberg’s recruitment website shows 42 open positions for quants specifically—with some roles paying up to nearly $300,000 per year—or for individuals to work on products targeted at quantitative analysts and traders, which the vendor sees as a growing area of demand for data, resulting in the creation of new products as quant strategies evolve.

This isn’t an AI story—though Bloomberg certainly knows its way around large language models—but I think it shows the changing nature of finance.

This whole column has just been a rehash of stories we’ve already published, but let me try to connect the dots. Firms aren’t necessarily looking for rainmaker portfolio managers and traders, but they want people that have something of a tech background to deliver better analytics and make employees more efficient.

That’s probably overly simplistic. But it feels like that the “old boys of Wall Street” are increasingly being replaced by tech and data specialists. If that is, in fact, true, then financial services cannot be the industry of “No, we can’t do that.” The struggle for talent has always been a headache, and not just for finance. But you do have to ask yourself if someone fresh out of college wants to work at an M&A investment bank or Bloomberg (or S&P or LSEG or SS&C or some other major tech vendor),or are they more likely to gravitate towards big tech companies such as Google or Amazon or Microsoft.

Data and analytics play into everything that drive the capital markets, and that requires top technologists and data professionals. But they also need to be part of the process and not default to the attitude of: “Hey IT! This system is down! WTF?!”

While generative AI and large language models are not necessarily new anymore, they largely are new in the capital markets. And as several of those people I’ve mentioned noted, if you don’t have people with an understanding of tech and data, it’s easy to get left in the dust when new technologies come onto the scene. And if you think that GenAI is the last major revolutionary example that we’ll see over the coming years, let me just remind you that the age of quantum will completely change everything.

I don’t know if I articulated this as well as I could have… it’s an idea that’s been bouncing around in my head. Think I can explain this trend more clearly? Or can you? Hit me up: anthony.malakian@infopro-digital.com.

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@waterstechnology.com or view our subscription options here: http://subscriptions.waterstechnology.com/subscribe

You are currently unable to copy this content. Please contact info@waterstechnology.com to find out more.

Most read articles loading...

You need to sign in to use this feature. If you don’t have a WatersTechnology account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here