Sign up to our newsletter

Data-driven insights and news
on how banks are adopting AI

UniCredit’s German payoff

3 October 2024

TODAY'S BRIEF

Some exciting news that we wanted to share. The 2024 Evident AI Index is just 14 days away… coming on October 17… and to mark the occasion we’re opening sign ups for the launch event. Join us for the Evident AI Index Roundtable on October 24 to dive into the new rankings and latest insights. And we’ll see you live at the Evident AI Symposium in NYC on November 21 – scroll down for our keynote speaker announcement and register today.

Now, in this week’s Brief, we dig into the UniCredit-Commerzbank deal, a bump in RAI patents, and Gen AI getting… cheaper? We’ve also got a great conversation with Wharton’s AI guru Ethan Mollick, a use case that feels kinda “big brother,” plus our top AI news and talent moves.

The Brief is 2313 words, a 7 minute read. If this newsletter was forwarded to you, subscribe here. We want to hear from you: [email protected].

– Alexandra Mousavizadeh & Annabel Ayles

TOP OF THE NEWS

ITALIAN JOB

The talk of the banking world is UniCredit’s “frenemy hug” of Commerzbank. The Italians’ move last week to raise its stake in the German bank to 21% from 9% is fueling speculation of a hostile takeover, potential leadership shake-ups, and a fresh wave of eurozone industry consolidation. It might also help these two banks close the gap on AI.

UniCredit and its “rainmaker” CEO Andrea Orcel are on a tear the past year, notching up the biggest stock price jump – over 70% – of the 50 banks in the Evident AI Index, 2.5x the financial sector’s performance as a whole, up 28.8%. But both Orcel’s bank and his target lag on AI adoption, placing 43rd and 42nd respectively in last year’s ranking of AI maturity.

Could this coupling, if consummated, change that? The Magic 8 Ball says… yes.

As we crunch the numbers ahead of the release of this year’s ranking (again, mark your calendars for October 17), we’re looking closely at why and how European banks have fallen so far behind the curve. In a word: Talent, or the lack thereof…

UniCredit illustrates the problem well. Over the past year, it made zero progress in attracting AI development staff. In fact, UniCredit was one of only two banks in the Index that saw no change in their AI development profile year-over-year (alongside Raiffeisen).

Now, if “UniCommerz” were to happen, that would be a game changer talent-wise. The combined AI development workforce would effectively rival that of Intesa Sanpaolo’s overnight. The two banks also complement one another: Commerzbank is stronger in Evident AI Index pillars AI Talent and Responsible AI, whereas UniCredit performs better in Innovation and Leadership.

'UniCommerz' Would Narrow Talent Gap

AI development staff at select European banks

Chart

Source: Evident Insights

There is precedent here: After UBS’s “forced marriage” to Credit Suisse last year, the combined AI staff from both banks were largely insulated from the massive (and ongoing) workforce reduction effort unfolding in Zurich.

WHAT’S ON AT EVIDENT

INDEX LAUNCH ROUNDTABLE

Event tile

Join us virtually at our upcoming Evident AI Index Roundtable on Thursday October 24 at 15.00 BST (10.00 EDT).

We will dive into the October 2024 Evident AI Index ranking to discuss:

  • Where are banks in their journey to AI maturity?
  • How has this changed in the past year?
  • Which banks are surging ahead and which are falling behind?

FROM THE EVIDENT AI INDEX

AI SAFETY AS A COMPETITIVE EDGE

Safety is a new new thing for the world’s top banks.

Nearly half the patents on RAI were published by the 50 Evident AI Index banks in the past 20 months, according to new Evident analysis. What’s more, we found 10 banks had filed a patent in Responsible AI so far in 2024, up from two in 2019.

Innovating Responsibly

Number of RAI patents published by the 50 Evident AI Index banks

Chart

Note: TTM = trailing 12-month period | Source: Google Patents, Evident Insights

What explains the push on RAI? That’s easy. AI is risky business and the topic is top of every Board pack.

The data shows that banks are becoming much less shy about talking about the risks: 38 out of 50 Index banks mention it in 2023 annual reports, up from 27 banks in 2022 (see “Risky Business”, The Brief, August 23).

Many banks perceive RAI as the one area where there is scope for collaboration in an otherwise highly competitive sector. We do see some go for shared information and joint partnerships, but the trend in opting to protect their safety measures viewing it as a competitive edge is certainly growing…

Strategy 1: Shared safety

Capital One entered academic partnerships with major institutions to develop joint research projects. BBVA shares open source XAI libraries with the wider community. RBC set up dedicated research hubs focused on responsible AI.

There are joint industry initiatives too, that recognize the value in sharing best practices on RAI.

Strategy 2: Competitive edge

A recent – and accelerating – trend is for major North American banks to patent systems that address bias and other externalities from AI implementation (i.e., black-box outputs, hallucinations).

JPMorgan last month patented a way to evaluate bias on AI/ML risk assessment models “in order to enable a large set of models to be reviewed efficiently.” And models can sometimes be so complex that you can’t understand how it got a certain result. The bank had to pull the plug on an FX deep learning algorithm for that reason, according to FX Markets.

More examples of safety patents include Bank of America’s expansion-based models to improve model fairness, and RBC’s adversarial attack model to diagnose model fairness.

Why does this matter?

Firms with a strong RAI department are better prepared to weather regulatory headwinds, to safeguard reputational risk, and to protect themselves against financial risks associated with unexpected model output.

The future of responsible AI in banking will likely see a contradictory combination of open sourcing and competition.

USE CASE CORNER

O BIG BROTHER, WHERE ART THOU?

In this week’s Use Case Corner, we look at three of the most interesting tools unveiled in the past couple of weeks.

And ICYMI: Check out our special edition Brief on 74 AI use cases.


#1 Employee compliance monitor

Use Case: AI-enabled employee communication monitoring
Vendor: Sedric
Bank: n/a

Why it’s interesting: Sedric has built a communications monitoring platform which uses LLMs to monitor employee communications (calls, chats, emails) to flag compliance issues.

Potential ROI → reduced compliance penalties, increased efficiency
Reported ROI → n/a


#2 Agentic AI moneymaker

Use Case: Consumer Agent
Vendor: n/a
Company: Bud Financial (“Bud”)

Why it’s interesting: Financial data platform Bud has added “agentic” capabilities to its consumer agent. Previously the agent could move money between accounts, and now it’s been trained to improve how much money a consumer earns in interest.

Potential ROI → increased customer satisfaction, income uplift
Reported ROI → For more than 27% of customers, the agent would have generated at least $500 in profit per customer over the course of a year, a study into the customer base of a U.S. bank found


#3 Scanning for scammers

Use Case: Fraud detection
Vendor: Mastercard
Banks: Halifax, Lloyds, Barclays, NatWest, TSB, Monzo, and others

Why it’s interesting: When a customer sends money to a recipient, Mastercard’s AI-enabled risk monitoring capability sends the customer’s bank a real-time risk score of the recipient account, flagging if it could be managed by scammers (“money mule” accounts). The capability adds to Mastercard’s growing arsenal of AI-powered fraud & security solutions.

Potential ROI → reduced fraud, increased customer satisfaction
Reported ROI → n/a

WHAT’S ON AT EVIDENT

HEITSENRETHER’S RETURN ENGAGEMENT

Speaker announcement

We are delighted to announce that JPMorganChase Chief Data and Analytics Officer Teresa Heitsenrether returns as our opening keynote speaker at our Evident AI Symposium on 21 November in New York City.

At the Symposium, we aim to cut through the AI hype in order to advance the global conversation of AI adoption in banking. Fuelled by the latest data from the Evident AI Index, we’ll discuss: where is the sector now – and what’s next?

Q&A

THE STATE OF AI ADOPTION

Ethan Mollick

Ethan Mollick, co-director of the Generative AI Labs at Wharton and the author most recently of Co-Intelligence: Living and Working with AI, spoke with Evident data scientist Alex Inch about all things AI: the pace of adoption, when experience matters, and how to foster experimentation. Here’s an edited and condensed version of their conversation.

INCH: Two years ago, you said generative AI is coming “faster than most people expect.” Looking back, how does that prediction hold up?

MOLLICK: Actual end user adoption has been insanely high, certain surveys that show 65% of marketers, 60% of coders are using Gen AI. But businesses don’t necessarily capture all of that value – that requires rethinking processes and approaches. If I increase everyone’s productivity by 20%... that’s awesome, but how do I as a firm actually collect on that? One way is to fire people, but if you fire people they won’t show you how they’re using AI.

How should banks get employees to experiment with AI?

You have to incentivize it. If employees are worried colleagues will lose respect for them for using AI, or the bank won’t reward them for using AI, or they will get let go because they’ve got AI to do 90% of their tasks, then they’ll never show you how they use it. AI access helps, sharing helps, some education/training helps, as well as showing support at the highest levels of the organization.

You’ve recently said that senior leaders that use AI are more effective at driving AI adoption internally. What use cases make sense for senior leaders if they want to dip their toes?

The research shows that juniors don’t know anything special about AI: they use it first, but they don’t learn specialized knowledge from using it. Seniors are actually the best able to use AI because they have the expertise, and they can get a sense very clearly of when the AI knows something or if it’s making it up. So the reason to use it is to understand what this thing actually does. If you just outsource that, you're not gonna get the right answers. JPMorgan's talked about how they have AI whisperers assigned to senior level executives, whose job it is to see how much of their job they can automate. But it’s less about helping senior executives, and more about them being hands-on with what this thing does.

You can read the full interview here.

SPEED READS

FIVE STORIES THAT CAUGHT OUR EYE

Goldman Sachs hosted a debate between two of its most prominent and contrasting voices on AI. Stock research chief Jim Covello made the bearish case while George Lee, co-head of Goldman’s geopolitical advisory business, stood his ground on AI optimism. And while summer is over, the next gen is sunny on AI: 93% of Goldman interns said AI is more likely to enhance (than replace) the work they do. Virtually all of them (99%) think regulation is essential.


Trust in AI chatbots is low, according to J.D. Power. While banks roll out the new tech (Erica from Bank of America, Fargo from Wells Fargo), consumers are skeptical – only 27% of Americans have faith in AI chatbots to give financial advice.


California governor Gavin Newsom vetoed a widely anticipated bill to regulate AI, prompting Alexandra to offer up her thought bubble:

  • Regulation is needed, but this was not the right bill. Instead of regulating the technology, we need to regulate the use of it. Banks have well-established governance structures working today that need to be adapted to specific AI risks, in coordination with regulators.

The Bank of England launched an AI Consortium for financial institutions to weigh in on AI regulation. To date, the UK has been hands-off on specific regulations for AI in financial services, but that appears to be changing.


Leaders at six Evident AI Index banks (BNY, JPMC, Lloyds, Truist, U.S. Bank and Wells Fargo) were featured among “tech execs to watch” in financial services for 2025. Recently appointed Lloyds group director of AI Rohit Dhawan filled a new role steering AI strategy.

TALENT MATTERS

EXITS AND BOOMERANGS

BMO appointed Kristin Milchanowski to the newly created role of chief AI and data officer. Milchanowski was previously at EY as global innovation partner focused on AI and is also an associate fellow in AI at the University of Oxford. The hope, according to CTO Steve Tennyson, is this new role will help the bank create AI products that differentiate it from competitors.


TD Bank named Raymond Chun its next CEO, succeeding Bharat Masrani on his retirement next April. Chung currently runs the personal banking business in Canada.


ABN Amro’s Chief Strategy & Innovation Officer Edwin van Bommel is leaving the bank. No news on where he’s headed next, but his departure post called out “great strides” on AI during his tenure, including an award for a call center AI solution.


Google acqui-hired legendary ex-employee Noam Shazeer, through his AI start-up Character. Shazeer agreed to rejoin Google to head up their Gemini model development.


Amazon is looking to Rohit Prasad, previously the chief architect of their Alexa assistant, to double down on the development of the company’s proprietary LLMs, which have struggled in performance and popularity compared to its peers.

MEET THE MODEL

GEN AI GETS CHEAPER

Google recently announced a 50% price cut for access to their LLM series Gemini. It’s a trend: ChatGPT creator OpenAI has cut the price of their GPT-3 model by 98% over the last 18 months. The cost of training a large language model may be exorbitant, but using one is getting cheaper.

More Models News:

  • Meta updated their Llama series with vision models that can see as well as read, alongside smaller text models that can run on a phone. Alas, these updates aren’t available in the EU, an increasingly common trend (see “World AI-nt flat”, The Brief, April 18)
  • OpenAI released its advanced voice mode to paying customers (again, outside the EU)... four months after the demo was accused of mimicking Scarlett Johansson.
  • Canadian startup Cohere collaborated with Fujitsu on a Japanese LLM. Many companies create localized models for a specific market, but Cohere appears to be the first to work on multiple markets with enterprise partners.
  • MIT spinoff Liquid AI released a new LLM architecture which outperforms traditional transformer-based models (i.e., the one everyone else uses). Hints of a new paradigm?

COMING UP

Tue 8 Oct
Bloomberg: The Business Value of AI, Los Angeles

Wed 9 Oct - Thurs 10 Oct
World Summit AI, Amsterdam

Thu 10 Oct - Fri 11 Oct
Collide 2024, Atlanta

Mon 21 Oct - Thurs 24 Oct
Sibos, Beijing

Thu 24 Oct
Evident AI Index Roundtable, Virtual

Thu 21 Nov
Evident AI Symposium, New York

THE BRIEF TEAM

Alexandra Mousavizadeh | Co-founder & CEO | [email protected]

Annabel Ayles | Co-founder & co-CEO | [email protected]

Colin Gilbert | VP, Intelligence | [email protected]

Andrew Haynes | Head of Data Science | [email protected]

Alex Inch | Data Scientist | [email protected]

Sam Meeson | AI Research Analyst | [email protected]

Mike Silverman | Director of Research | [email protected]

Matthew Kaminski | Senior Advisor | [email protected]