AI-leaning banks kill on earnings

Good times keep rolling. UBS (#6 in the latest Evident AI Index) this week smashed profit expectations, while HSBC (#7) saw quarterly profits jump 10%. Europe’s big players followed the hot results put out by the Americans.

Many of the beats on earnings are dizzying (see below), led by Goldman Sachs’s 23.4% improvement on its forecast and Morgan Stanley (18.2%), whose stocks are hovering at all-time highs.

What’s going on: Basically, interest rates are down, deal fees up. The banks talk about ROI on AI less than their tech peers (driving Alphabet up and Samsung down), but that’s bound to change, if you listen to what the CEOs are saying this quarter.

• BNY’s Robin Vince: “We haven’t made a lot of noise about it, but don’t misunderstand that for a lack of interest or investment because we haven’t slowed down. In fact, we’ve increased our AI investment.”
• UBS’s Sergio Ermotti addressed AI at length, citing the largest deployment of Microsoft Copilot within the global financial services (50k licenses) and expansion of the bank’s proprietary AI assistant (“Red”) to 20k users operating in three markets. “We are preparing for the future,” he said, “this includes investing in our industry-leading cloud infrastructure as well as our expertise in artificial intelligence and automation. This will accelerate generative AI adoption, increasing efficiency and effectiveness.”

Additional banks hosting analyst calls this morning include: #12 BNP Paribas, #13 BBVA, #17 ING, #21 Société Générale, #23 Intesa Sanpaolo, #41 CaixaBank, and #42 Danske Bank.

This results season confirms a trend. The banks that lean into AI tend to outperform their peers on share price – as we showed last spring in looking over 12-month stock performance (see “Market Rewards Evident AI Leaders,” The Brief, April 18) – and now we can see that with earnings and revenues.

The average Index bank is currently beating earnings per share targets by 7.1% and revenue targets by 2.6% (see chart above) – that’s the difference between the consensus of analyst forecasts and the quarterly result, in technical jargon the “surprise”. But as you see below, the banks in the top half of the index are trouncing counterparts in the bottom by 2x on EPS and 3x on revenue.

What it means: While the Evident AI Index could provide a leading indicator for financial performance, we need to be careful to distinguish between correlation and causation. Are banks that exhibit higher AI maturity performing better? Or are banks that are performing better overinvesting in AI? Time will tell… along with full fiscal year results.

The Evident AI Symposium – an unmissable gathering of the top minds shaping AI in banking.

This is your chance to get a first hand look at what is going to shape AI adoption in 2025.

Following October’s Evident AI Index update, this exclusive event will dive deep into which banks are leading the way? Where are they delivering value? And what are the headwinds and opportunities on the horizon?

View the latest agenda and recently confirmed speakers here

We look forward to seeing you - either in person, or virtually.

In this week’s Use Case Corner, we present the three most interesting AI applications that banks rolled out in October.

#1 GPT research assistant

Use Case: Process Automation and Knowledge Access
Vendor: Internal development with OpenAI's GPT-4
Bank: Morgan Stanley

Why it’s interesting: AskResearchGPT gives Morgan Stanley’s investment banking, sales & trading, and research teams access to summaries of insights from over 70,000 annual research reports instantly.

Potential ROI: Improved productivity with faster research access, better client management
Reported ROI: n/a

#2 AI Assistant for Life Insurance Underwriting

Use Case: Process Automation and Knowledge Access
Vendor: Microsoft
Bank: BMO

Why it’s interesting: Using Microsoft’s OpenAI Azure service, the assistant gives advisors instant, automated access to essential field data for underwriting that usually requires sifting through numerous documents. Advisors can respond to client questions and search documents faster, all in real-time.

Potential ROI → Reduced underwriting time and operational costs
Reported ROI → n/a

#3 Multilingual Regulatory Compliance Assistant

Use Case: Regulatory Compliance
Vendor: n/a
Bank: Deutsche Bank

Why it’s interesting: “Project Aggie,” still in the demo phase, is a multilingual regulatory assistant that helps streamline compliance and regulatory requirements across jurisdictions and languages. What typically took hours of expert input now might be done in under five minutes using generative AI.

Potential ROI → Increased compliance efficiency
Reported ROI → n/a

Have feedback on or ideas about use cases? Let us know at [email protected].

"I can’t think of a wealth management client today that is not coming to us to say ‘what do we need to do, how do we get started?’"

- Sarah Friar, OpenAI CFO, interviewed by Ed Ludlow at Money 20/20, Oct 28

Intesa Sanpaolo plans to cut 9,000 positions in the next three years to push a digital transformation and “generational shift.” It will encourage pension-ready employees to hit the exits voluntarily, hire 3,500 young people and focus on digital and AI.

Bank of America reported its portfolio of AI and ML patents grew by 94% since 2022. The bank is tops in AI patent citations, according to Evident’s most recent data.

AI puts a strain not just on energy grids but also our water supplies, according to a report from JPMC and ERM. Data centers are using lots more of it to cool down their hardware.

New York State’s Department of Finance proposed guidelines on cybersecurity risks related to AI. It recommended financial firms factor AI into their risk assessments, conduct due diligence on vendors and train staff on AI and cybersecurity. No brainers to most banks, but shows that regulators are waking up to AI.

The White House asked the U.S. AI Safety Institute to give guidance to American AI developers on how to test and manage foundational model risks. It set a 180 day timeframe on this, but isn’t pushing mandatory rules yet, suggesting a hands off approach here no matter who wins next week. (See Q&A and Coda below for more).

Has responsible AI (RAI) found its Blake-ian “green and pleasant land” in the UK? By almost all measures of RAI in the Evident AI Index – such as publishing ethical AI principles, adapting risk management frameworks or mentioning RAI in strategy documents – British banks outperform their peers.

Most striking this year is how UK banks have bolstered their RAI talent ranks: Headcount has effectively doubled in the last year, representing the biggest shift in resourcing across all six geographies.

One driver of this is how UK banks have prioritized leadership roles for RAI. Earlier this year Natwest appointed ethical AI guru Paul Dongha its head of RAI & AI strategy. Lloyds hired Rohit Dhawan (ex-Amazon) to head up a new AI Centre of Excellence that brings together the AI ethics team with data science and ML staff.

P.S.: JPMC is hiring an AI/ML Governance lead, overseeing AI risk management and impact assessments across multiple projects in the Corporate & Investment Bank.

Any noteworthy people moves you know about? Share them with us at [email protected]

And one last note: We cover the latest rankings for RAI in the latest Evident AI Index, but will be going deeper on the data in the Transparency Pillar report, going live in January 2025 - stay tuned.

Democratic congressman Ro Khanna from California, and Republican congresswoman Kat Cammack from Florida spoke to Evident Senior Advisor and POLITICO Editor-at-Large Matthew Kaminski at Deel’s Future of Work Policy Summit on Oct 24 about the U.S. elections, policymaking and AI.

Here’s a shortened, edited version of their conversation.

KAMINSKI: What should the next U.S. Congress and federal government do on AI regulation?

CAMMACK: As AI technology has really become a part of so many different sectors and continues to have a really important role moving forward, you get very nervous about the big players in the space coming in with a set of regulatory recommendations which ultimately would hamper innovation and push out smaller players in the space. So from my standpoint, we should focus on the foundational elements, the language models, using blockchain as a solution for credentialing and providence credentialing.

KHANNA: In my district in Silicon Valley, we have $10 trillion of market value. We're gonna have probably another $10 to $20 trillion of value generated in AI. I'm very, very optimistic about the future of economic innovation. But the question is, is that economic gain going to be concentrated in places like Silicon Valley, or are ordinary Americans going to benefit from it? And a lot of policy decisions will go into that. The tax code has implications there. Right now, what you have is companies can deduct for a capital expenditure, but they don't have the same ability to get tax credits for hiring people or investing in people.

CAMMACK: I doubt that you'll find that Republicans will be in favor of any additional taxes.

Where do you two agree on policy solutions as it relates to AI?

KHANNA: I'm hopeful, of course, that Kamala Harris will be president. But if Trump becomes president, he has Elon Musk on his side. I know Elon very well. Elon and I did a chat on X with Mike Gallagher. He was no longer in Congress, but [he is] a Republican. And Elon proposed there and in other places having an AI regulatory commission, a federal agency on AI to make sure that we had basic safety tests, to make sure that we had clear regulation in terms of something being AI or not. I would support an idea like that.

KAMMACK: Regardless if you're an R or a D, we have to have a national data privacy standard. Because the piecemeal work that has been done state by state, it's just not sustainable. No company has the bandwidth to comply with 50 different states data privacy regulations.

If 2023 was the year of AI hype, the balloon shriveled a bit this year. When do you think AI will really show concrete impacts?

KHANNA: I think AI is already having an extraordinary impact. It's having an extraordinary impact on our military. AI is having an impact on medicine already in terms of doctors and nurses being able to use it. The person who won the Nobel Prize at Google had AI help discover new proteins. We're gonna be having AI helping in new educational platforms. Now, in terms of gross displacement of the labor force, my view is that the production capability of America will expand dramatically so that we don't have a mass displacement of workers. But what we need is to figure out how we make sure everyone is participating in this prosperity.

How concerned are you that we’ll still see some kind of AI surprise in this election?

CAMMACK: Everyone was freaking out about the deep fakes. While AI is good, it's not that good. People can still pick up on it by and large. I don't think we're going to have that. I'm feeling pretty confident.

KHANNA: I do think that you had AI enable bots to proliferate propaganda. The deeper problem is we probably need more people learning kindness, respect from kindergarten to sixth grade and civic education, more than thinking we're going to solve this by dealing with bots.

The striking thing about AI and the U.S. election has been – aside from one notable deepfake endorsement from Taylor Swift – the absence of AI throughout this campaign. The most consequential disruptor since the Internet came up exactly zero times in the two presidential debates. The Kamala Harris website mentions AI four times, Donald Trump’s eight.

But as the Q&A above shows, there are clear enough differences between Red and Blue in America, both ideological and practical, to be able to sketch out what a Harris and Trump administration might mean for AI.

Steady Harris Goes

Between Biden’s Executive Order from October 2023 and the National Security Memorandum (see summary in Speed Reads) from last Friday, President Harris would come in with an established policy foundation. As the former “AI Czar,” Harris helped orchestrate voluntary safety standards that were signed by OpenAI, Microsoft, Google, and Anthropic. She also championed her predecessor’s Executive Order at the UK AI Safety Summit last year.

The Biden administration hit the antitrust pedal against Big Tech. If self-styled tech trustbuster Lina Khan stays on at the Federal Trade Commission, expect even more of that, which won’t be received warmly by the big players in Silicon Valley.

Trump Laissez AI Faire

A Trump administration would likely be less bolshy on the regulatory and antitrust fronts. While Trump signed Executive Orders on AI in February 2019 and December 2020, the 2024 Republican platform called Biden’s EO “dangerous” and promised to repeal it, arguing it hinders innovation. Trump’s running mate J.D. Vance has warned that “presumptive overregulation attempts… would frankly entrench the tech incumbents that we already have.”

As Trump gained momentum this past month, business leaders (including Big Tech) reached out to the Trump campaign presumably to hedge their bets going into next week. Beyond the early movers into his camp (see Elon Musk, Marc Andreessen, etc.), this reinforces the simple truth that Silicon Valley skews more purple than blue.

Consistent, They Are Not

Both candidates and parties have cut against their default positions. California Governor Gavin Newsom, a leading Democrat, recently vetoed California’s AI Safety Bill (SB 1047), which OpenAI, Meta, Y Combinator, and Andreessen Horowitz actively lobbied against. Harris hails from California and is close to Newsom, which leads some to speculate if she’ll bring “Biden AI policies with a Silicon Valley-informed approach?”

For his part, Trump has threatened to slap tariffs on the Taiwanese chips that power American tech. Given the lack of U.S. semiconductor fabs to keep pace with the scale of AI investments, that could curb plans for all those nuclear-powered AI data centers and end up hindering innovation.

A last reality check: Washington is the analog capital of the country, and policymakers have never been able to keep pace with fast-moving technology. Why would they start now?

Do you know or run an event that you think should be featured? Let us know at [email protected]