In November 2023, Evident convened 140 of the most influential global AI leaders, technologists, academics, and policymakers in New York City for our inaugural AI Symposium. This event was one of the largest gatherings to date of individuals tasked with accelerating the adoption—and ultimately, the impact— of AI across the banking sector.
A total of 82 companies had senior representatives attend in person, with 25 global banks taking part. 45 of the 140 attendees were either President, Chairman, C-Suite, or MDs.
Our media partner Bloomberg and our partners EY, Deepsee, and Cognaize helped us bring this exciting event to fruition.
We heard from many exceptional speakers including Clare Barclay, CEO at Microsoft (UK); Satvinder Singh, Global Data & Analytics Division Lead at LSEG; George Lee, Co-Head of the Office of Applied Innovation at Goldman Sachs and Nimrod Barak, Global Head of Innovation Labs at Citi.
At each session on the day’s agenda, speakers emphasised the scale, speed and importance of AI adoption efforts throughout 2023.
Responding to a question about the banking industry’s relative maturity as compared to other sectors, Clare Barclay from Microsoft made this vital point: "I've seen, if anything, the most progression in thinking, adoption, strategy, and piloting in Financial Services—of all the industries I've worked in. The challenge remains turning a proof-of-concept into production."
Teresa Heitsenrether from JPMorgan Chase stressed the importance of committing long-term to AI adoption: “In the early years, you don’t necessarily see the results immediately. A factor of our success is our commitment to the investment—and not just investing in the latest technology, but all of the foundational pieces you need—the data infrastructure, the cloud computing, the embedded governance components that make this all work as one ecosystem.”
From the left: Caroline Hyde (Bloomberg) & Teresa Heitsenrether (CDAO, JPMorgan Chase) gives a 'Keynote Speech'
Coming away from these forward-looking, thought-provoking discussions with 20+ experts shaping AI transformation across the banking industry, several common common themes emerged from the agenda that are going to shape our predictions for the coming year:
- Generative AI Accelerates from Promise to Reality
- AI Leaders Expand Presence on Operating Committees
- AI Talent Escapes Right Sizing & Cost Reduction
- Banks shift from: What Can AI Do → What ROI Does AI Create?
- Banks provide the test case for emerging public policy on AI
- The gap between leaders vs. laggards widens
1. Generative AI Accelerates from Promise to Reality
Generative AI is often cited as the be-all, end-all solution for banks seeking to enhance fraud detection, automate customer service, and optimise risk assessment. This application is consistently identified as the key to unlocking future growth potential. In the words of David Rice (HSBC): "If you are really going to have a transformation in your current customer relations, then you'll need to complement traditional AI with Generative AI to drive non-linear growth."
Despite that, many speakers stressed that customer-centric use cases are a ways down the road. Foteini Agrafioti (Royal Bank of Canada) stressed: “Just to focus on Generative AI—we do not feel, like many others, that the technology is ready for primetime, if I define primetime as client-facing in the near-term. We don’t expect to see a client interacting with a chatbot to get financial advice in 2024.”
In fact, only one speaker specified a “fully live” solution based on an external large language model. Jeff McMillan (Morgan Stanley) emphasised: “First of all, you’re not going to build these things [LLMs] yourself—at least, not this year. We’re the only financial services firm in the world with a direct partnership with OpenAI’s research team. Two years ago, that was the only firm with an offering in this space.”
Projected time horizons on when we might see GenAI models move from proof-of-concept to production across other banks vary widely. Monique Shivanandan (HSBC) suggested this could occur in the next 3-6 months, whilst Stefan Simon (Deutsche Bank) estimated it might take 12-18 months. Clearly, ambitions for scope and implementation depend on the geography and regulatory environment of a given bank.
While there are competing estimates regarding when, broader consensus emerged on how banks would engage in a phased shift from preparing for GenAI deployments to expanding access to internal, private tools that look and feel akin to public versions of ChatGPT.
Basically, the general order of operations is as follows. First, we’ll see GenAI use cases deployed to back office and middle office functions (a.k.a. “no regret” applications). Second, we’ll see internal, private chatbots deployed to front-line workers to help speed document discovery and summarization. Third, we’ll see GenAI deployed to accelerate code generation, which speakers frequently stressed should be in tandem with established safety checks and human review. As Clare Barclay (Microsoft) emphasised: “ . . . we believe in a fundamental principle of co-pilot and not autopilot.” Other potential use cases—such as risk modelling of liquidity positions—remain a long way out.
David Rice (HSBC) summed it up best: “Banks are an information services business, so the possibilities here . . . the opportunities here . . . are nearly endless. But also the level of disruption and transformation. So the opportunities are both endless in one respect, but constrained in terms of how we methodically change our infrastructure to do this at scale. The time to get to mass adoption is how fast you safely scale that infrastructure.”
From the left: Clare Barclay (CEO, Microsoft UK) & David Rice (Global COO of Commercial Banking, HSBC) talk on 'The Long Game'
2. AI Leaders Expand Presence on Operating Committees
In 2024, we expect to see more banks acknowledge the importance of AI at the executive level–either through existing team members driving the bank’s AI agenda through both internal and external channels, or by appointing new Operating Committee members with a dedicated AI/Data remit.
Across the 50 banks Evident tracks, we found only three instances where a Chief Data & Analytics Officer (CDAO) had been elevated to the Operating and/or Executive Committee (JPMorgan Chase, NAB, and CaixaBank). In addition, we have found no evidence of a Chief AI Officer (CAIO) across the Index, following Tomi Poutanen’s departure from TD Bank in April 2022. We expect this trend to become more pervasive, especially with the latest White House Executive Order (14110) requiring Federal agencies to designate a CAIO within the next 60 days (see subsection 10.1b).
This simple, but impactful change to executive leadership teams has newfound urgency. Evident data shows that only 16 banks have at least one Board Member with existing AI experience from their tenure at a previous company. As Kris Pederson (EY) hammered home: "The board needs to understand the technology it is tasked to govern. The background of the board matters. Demystifying these technologies for the Board Directors is critical."
From the left: Eric Schatzker (Bloomberg), Kris Pederson (Americas Center for Board Matters Leader, EY), Monique Shivanandan (CDAO, HSBC) & Stefan Simon (CAO & Head of the Americas, Deutsche Bank) talk on 'The Compliance Question'
3. AI Talent Escapes Right Sizing & Cost Reduction
The demand for AI talent within the banking sector is surging, with associated headcount for AI Development, Model Risk, Data Engineering, and Implementation up +9.6% across the world’s biggest banks from May 2023 to Sep 2023 (despite a -2.5% reduction in overall headcount). This pattern holds true even in the aftermath of major mergers and consolidations, such as that between UBS and Credit Suisse which was finalised this past June.
Despite insulation from broader cost cutting measures, banks are not only in fierce competition with each other for scarce talent—but with established partners. Over the last two years, five of the top 10 “poachers” of AI talent from banks have been tech companies (e.g., Amazon, Microsoft, Google, IBM, Oracle)—collectively responsible for a third of defections from banking to other industries.
Second only to the primary topic of the day (AI), “talent” was the most mentioned watchword throughout Wednesday’s programme. Specifically, talent was referenced as an ongoing and systemic challenge that cannot be addressed overnight. Commenting on her bank’s strength in this area, Teresa Heitsenrether (JPMorgan Chase) reflected: “I think you have to go back a little bit in time, because it's certainly not something that happened overnight. I think we did a lot of things that were very smart—we brought in some super talented people, we created a lot of connections with academia, we did a lot of branding exercises, we did a lot of training programmes . . . we created the flywheel. And I think what slowly happened over time is that people understand you have incredibly interesting problems to solve, we have vast amounts of data, and we have the ability to invest. So if that’s your passion, what better place to work than to come to a place where you can perfect your craft and work in an environment where you’re going to continue to be on the cutting edge?”
4. Banks shift from: What Can AI Do → What ROI Does AI Create?
While 39 out of 50 banks now reference AI somewhere across their strategic documents, the depth and specificity with which they do so varies dramatically. In fact, only four banks (JPMorgan Chase, DBS Bank, Société Générale, and BNP Paribas) articulate the total number of AI use cases in production at the bank, as well as the expected ROI.
Over the past year, several banks have already seen a stock price bump based on their perceived affinity with AI, ramping up the pressure on senior leaders who aren’t as well-versed in the bank’s AI activities to engage more frequently and in greater detail on this topic than ever before. In many cases, this requires leaders to establish how far their institution has progressed in the overall journey.
As Teresa Heitsenrether (JPMorgan Chase) stated upfront: “In the early years, you don’t necessarily see the results immediately. A factor of our success is our commitment to the investment—and not just investing in the latest technology, but all of the foundational pieces you need—the data infrastructure, the cloud computing, the embedded governance components that make this all work as one ecosystem. All of that foundation has to be in place to be successful. But I think the other interesting thing that we’re starting to see now is that in addition to operational efficiency and cost avoidance—we’re seeing revenue-generating activity, which is very encouraging.”
Without question, more and more scrutiny will be brought to bear on persistent questions regarding ROI. How does your AI strategy differ from the immediate competition? How much is your bank investing in AI? What initial returns are you seeing from these projects? For some, this could blunt the experimentation and innovation needed to get an AI strategy off the ground.
While many stressed patience when it comes to recognizing (and reporting) near-term returns from AI investments, several contributors stressed a shift from experimentation to monetization. Deborah Lindaway (PNC) shared how the cost-benefit analysis of potential AI applications is changing: "One approach we're taking is looking at how these business use cases get elevated at the firm so that they are consistent. It’s no longer going to be someone’s pet project because they talked to a new vendor. Bringing that standardisation together, but also building that ROI, self-funded use case. What you have to balance that against is not dampening the creativity and ideation. But if it’s not part of a normal business-as-usual or technology line item in the budget, you need to put forward a solid business case and get very senior-level buy-in up to the Executive Committee level, if needed.”
When asked point blank about what metrics he can share, Jeff McMillan’s (Morgan Stanley) response drew sympathetic laughter from the crowd: "We didn't put economic KPIs against early GenAI pilots, because the goal was learning. But there's meaningful savings for all of us in these tools. It's not hyperbole. It's real."
Jeff McMillan (Head of Analytics, Data, and Innovation, Morgan Stanley Wealth Management) talks on 'The Collaboration Imperative'
5. Banks provide the test case for emerging public policy on AI
AI risk and regulation have entered the mainstream. From the UK AI Safety Summit, to the White House Executive Order to the EU AI Act, a range of concerns have surfaced—spanning long-term existential threats and geopolitical risk to more immediate issues around model bias and surveillance. Meanwhile, the OECD AI Incident Monitor (AIM) is reporting an acceleration of reported harms from AI, providing compelling evidence that effective regulation is not only necessary, but essential.
In this multipolar environment, there’s an outstanding question on what role banks can play as different regulatory frameworks move forward at different rates. Distilling the problem into fundamental dynamics, Stefan Simon (Deutsche Bank) observed: "Banks are on the receiving end. We are not managing our regulators and supervisors, but we can engage with them and explain the technology—especially where there is broad consensus on the principles of AI."
Arguably, as a highly regulated industry, banking is better positioned to influence (versus react) to these policy initiatives by promoting effective risk frameworks that have been carefully reviewed and implemented throughout the past year—specifically NIST’s AI Risk Management Framework 1.0 (January 2023) and ISO/IEC 23894 (February 2023).
Expressing confidence in the industry’s ability to meet and exceed emerging requirements from competing regulatory agencies, Forteini Agrafioti (RBC) stressed: "I don't go to bed at night worried about unintended consequences. I trust our protocols. I trust our controls. And I feel the wider industry shares our same level of commitment."
And pragmatically (as always), Manuela Veloso (JPMorgan Chase) suggested that while the technology at play is different, banking has several models to follow: "Regulation will become a testing goal—same as for food or drugs. The business will tell us when the outcome is wrong and we will re-do, re-test, get more data, get more input . . . until the AI passes the test."
From the left: Eric Schatzker (Bloomberg), Manuela Veloso (Head of AI Research, JPMorgan Chase), Stephen Flaherty (CTO, Barclays) & Nimrod Barak (Global Head of Innovation Labs, Citi) talk on 'The Leading Edge'
6. The gap between leaders vs. laggards widens
Only three banks (JPMorgan Chase, Capital One, and Royal Bank of Canada) score >50 of 100 points in the Evident AI Index—establishing a decisive lead over the rest of the industry in terms of their relative AI maturity.
Based on this warning shot, we’re seeing widespread urgency across all banks in the Index, coupled with careful consideration of where and how each bank can differentiate itself. As the barriers to entry for AI deployments continue to decrease, the volume and diversity of use cases in development may no longer be the only objective.
Going forward, identifying unique, proprietary use cases specific to each bank (versus what’s defined as “table stakes” for the wider industry) will become critical in ongoing communications with investors, customers, and prospective hires. Per George Lee (Goldman Sachs): "Where do you have data that is separate and differentiated? Where do you have data that's different from everyone else—and how do you use the machine to express that advantage? It's a different way of thinking about the world…”
As banks enter this critical phase in the continuing race for AI maturity, few are equipped with the foundational elements to accelerate as fast as they would like—or at the continuing pace set by the Index leaders. Here, several speakers commented on the limits of “first-mover” advantage—and the opportunities that exist to rapidly close the gap with established leaders.
Per David Rice’s (HSBC) argument: "Because of the open source origin of many AI technologies, there is little competitive advantage in the underlying tech—the competitive advantage resides in your proprietary data. Data is such a valuable asset, it should be on your balance sheet."
Vahe Andonians (Cognaize) echoed this sentiment: “The last 20 years were about algorithms. It’s going back to data and to capitalise on that data. Banks are in a good position here. I think the big difference here is there are markets . . . where the second mouse gets the cheese. It’s not so much about being first with something, because you can protect it.”
In response, Deborah Lindway (PNC) capitalised on the analogy: “We will not be first [to market] . . . we will be that second mouse that gets the cheese.”
From the left: Deborah Lindway (Executive VP, Enterprise Technology & Security, PNC), Vahe Andonians (CTO, CPO, Founder, Cognaize) & George Lee (Co-Head of the Office of Applied Innovation, Goldman Sachs) talk on 'The Generative Revolution'
Regardless of starting position, current capabilities, and investments made to date—one thing is clear. Inaction on AI is not a viable option and comes at a high cost to growth, productivity, financial performance—and increasingly—relevance in a rapidly changing industry.
—
Thanks to all our partners, speakers, and attendees for all your contributions to the events and the continuing dialogue.
Alexandra Mousavizadeh
Annabel Ayles
