Artificial intelligence

Banks may be eager to take advantage of opportunities posed by artificial intelligence, but they have work to do in persuading their customers to get onboard.

As they weave AI into their operations, seeking to make customer-facing and in-house functions more efficient, banks “really need to amp up their transparency and the education that they're providing customers,” said Ila Ghosh, senior director of financial services intelligence at J.D. Power.  

Less than 30% of consumers trust AI chatbots for financial information and advice, and less than half believe AI will improve their personal finances, according to a J.D. Power report released in September.

Personalized alerts and recommendations hold some appeal to customers, “but many are wary of allowing AI to control the actions and the application of personal security measures,” J.D. Power said. The firm surveyed about 2,000 U.S. financial services customers on their AI use and opinions in April.

Roughly 54% of bank customers said they’ve used some form of genAI, and that’s more likely among big-bank customers than regional ones, J.D. Power said. Banking is predicted to be heavily affected by generative AI, and with the promise of efficiency gains, the industry has scrambled to explore uses to bolster efficiency and productivity. 

Although a hot topic, AI remains somewhat of an enigma, at least to consumers, Ghosh noted. Only about one-third of customers have a solid grasp on how AI works, and many can’t tell when AI is being used, she said.

“The onus is on financial institutions to communicate to their customers a clear understanding of how AI works and its benefits,” J.D. Power said in the survey findings.

Ghosh was struck by consumers’ strong reservations around AI’s role in managing personal finances, despite its current integration into the financial services world through chatbots and fraud detection. 

Bank customers expressed apprehension around data privacy and security, increased risk of fraud and loss of human support when AI is involved. These customer concerns aren’t unfounded when AI is evolving rapidly, Ghosh said.

There’s also unease about potential bias and fairness issues related to AI, particularly regarding lending decisions, said David Donovan, executive vice president of financial services for the Americas at consulting firm Publicis Sapient. 

Room for optimism

Banks have started to leverage AI in substantial ways, including by offering customers tailored financial advice and quick assistance through chatbots and virtual assistants, Donovan said. As for banks’ own operations, AI is automating tasks such as fraud detection, payment processing and loan approvals, he said. 

As interest in generative AI has increased, banks are pursuing its use to help software engineers develop code, or make virtual assistants smarter and more efficient. Publicis Sapient’s recent study polled 1,000 senior banking executives and discovered about one-third of banks’ customer experience investments are going toward AI, machine learning and genAI.

Customer perception of AI, however, “is a bit of a mixed bag at the moment,” Donovan said in an email.

Most consumers believe banks are using AI to prioritize their own interests over those of customers, Ghosh said.

Of customer-facing ways banks are tapping AI, consumers expressed the most interest in alerts that might offer them a tangible benefit: Close to two-thirds of bank customers are interested in trying AI-powered account alerts that would help them avoid service charges and fees. 

Consumers were less enthusiastic, however, about using automated phone voice assistants to handle customer service issues (44%), or taking personalized product recommendations generated by AI (42%). 

On the customer service front, customers are worried it’ll be harder to reach a human when they need to address complex issues. 

“People want to talk to people. Machines can’t do everything,” Ghosh said. 

In light of that, it’s not surprising that consumers want their banks to employ a significant level of human oversight when AI is used, she said.

Customers also want to be informed when AI is involved in their interactions with banks. 

“Not every interaction may require disclosures, especially when it’s enhancing security or providing a seamless experience, but I think the customer perception is that they want to know if a bank is using AI,” Ghosh said.

Enhanced transparency  

There’s a clear need for banks to be more transparent about their AI use and provide disclosures, particularly as AI becomes more ubiquitous in banking, analysts said. De-mystifying the technology, by explaining how their AI systems work and what their data protection practices are via demonstrations on their websites, can benefit their relationships with customers, Ghosh said. 

“Customers deserve to know exactly how their data is being used, how these AI-powered decisions are being made, and what safeguards are in place,” Donovan said. 

Banks should emphasize AI’s security benefits, whether that’s enhancing fraud detection or real-time transaction monitoring, and offer clear explanations on how AI processes data, he said. 

Banks also need to give customers visibility into how AI decisions are made, Donovan said. 

“This could include customer-facing dashboards showing the logic behind loan approvals or investment recommendations,” he said. “Banks should be leveraging every channel – like emails, pop-up notifications and dedicated web portals – to educate and inform their customers.”

Lenders should also ensure their customer service representatives are educated on the bank’s AI use, Ghosh said.

There could be a need for targeted AI education to reach demographics with a wary attitude toward AI, she said. College graduates and customers who are men or under 40 use AI to a greater degree and have a more optimistic outlook on the technology than do women or older consumers, J.D. Power found.

Building customer confidence in AI's ability to handle complex financial matters is a “critical challenge” for banks, Ghosh said. But “anything that can help customers understand how the bank is using AI, why they’re using AI and really how it’s going to benefit the customer” can aid in efforts to alleviate customer fears and build trust, she added.

Banks may be enticed by short-term gains that AI can offer efficiency and personalization, but lenders have to take a long-term view, Donovan said.

“Banks have to double down on demonstrating the ethical, unbiased use of AI,” he said. 

Partnering with independent third-party auditors to validate banks’ AI practices as beneficial could be a smart move, Donovan added. 

“And proactively communicating their stance on fairness and transparency is key,” he said. “Customers want to feel in control, so giving them the ability to opt in or out of AI-driven services is crucial.”

Michael Ruttledge has no doubt Citizens Bank is merely scratching the surface when it comes to the use of generative artificial intelligence.

But the $219.9 billion-asset bank’s pacing in its use of the burgeoning technology is key, keeping security and customer privacy top of mind, said Ruttledge, chief information officer and head of technology services at Providence, Rhode Island-based Citizens, during an interview in July.

“We're not rushing into this,” Ruttledge said. “But I know by the pent-up demand … by the number of use cases that people want to go forward with, that the ideas are there, the demand is there.”

Citizens recently completed a pilot that tapped genAI to help the bank’s software engineers develop code, said Ruttledge, who was a unit CIO at American Express before joining Citizens. Now, the bank looks to roll out the capability to about 1,000 software engineers this year and ramp up through next year. 

Editor’s note: This interview has been edited for clarity and brevity. 

BANKING DIVE: Citizens CEO Bruce Van Saun has cited strategy as a top challenge for banks when it comes to employing AI. Where is the bank at with genAI ideation versus implementation?

MICHAEL RUTTLEDGE: We’re putting a very deliberate, dominant structure around the use of genAI. We started with basically a hackathon: we generated over 90 different ideas across the bank, and we honed it down to half a dozen we were really going to pursue. Two are actually in production. 

One is the use of Microsoft Copilot, which our engineers can use to help develop code. We’ve just wrapped up a pilot with a group of engineers, which we saw a 20% productivity boost from. We’re going to be rolling out Copilot to about 1,000 of our engineers. People are using it to do core code conversions, taking SAS code and converting it to Python. They’re using it to do documentation, they’re using it to understand legacy code that maybe is not well documented. They’re using it to generate test use cases. It’s an extension of what we’ve been doing, because we’ve really been looking to enhance the productivity of engineers. 

The second is around our contact center, and we developed a genAI model that replaces a legacy app that our contact center representatives can use when a customer calls in and asks a product question. It will come back with an answer about what the best product is to offer that customer, or that product’s features. It’s a colleague satisfier and a customer satisfier. Call resolution time has gone down. These two use cases are leveraging the same platform, rather than having everybody develop their own solutions. 

We have taken a deliberate approach, probably over the last six, nine months, because we had to put the appropriate guardrails in place, the appropriate security and governance. So throughout these use cases, we have the human in the loop, so there's someone validating the code as written before it's launched in production. There's someone validating and checking the answers before we give it to our customers. We haven't done anything yet direct to consumer.

When do you expect that might happen for Citizens?

Eventually, that will definitely happen. You can take a risk lens on it, and there’s probably certain use cases that are more likely than others. Ultimately, chatbot technology, you’ll be doing it direct to consumers. 

My personal opinion on that? Probably in a year or so, for us. Not for this year.   

What else has been crucial as the bank evolved its approach to technology? 

Our next-generation technology strategy really moved the bank to a more agile operating model, and brought in a lot of the tools and processes you need to do that, such as DevSecOps (development, security and operations), pipelines, test automation, building APIs, moving to the cloud and beefing up our engineering talent at the bank. 

When you think about the cloud, previously it would take us months to provision infrastructure: we’d order a server, install the server; now that’s all automated. And we’ve built in controls along the way. It’s a huge time-saver. We’ve also done a massive amount of upskilling. We’ve reskilled about 600 of our engineers over the course of three, four years — reskilling them in the cloud, modern technology, more recently in data engineering and genAI — and we’ve hired upward of 750 software engineers.

Where is the bank at with its move to the cloud? 

We will be completely out of all our data centers by the end of next year. We just crossed a major milestone in mid-July, in that we now have more than 50% of our applications in the cloud. We have a hybrid cloud strategy with Amazon Web Services and Microsoft Azure. By the end of this year, we’ll be at 70%, and by the end of next year, 100%.

During a congressional committee hearing earlier [in July], Federal Reserve Chair Jerome Powell mentioned cyber and an attack on a major bank as a persistent worry. What are some of your biggest concerns?

Cyber and stability are always something we heavily focus on. We’ve got a really top-notch team to focus on that. With the world situation right now, nation-state threats are always there. It’s constantly top of mind, but we’re very well prepared. I have monthly briefings with my team on these events, we meet with the board on a regular basis, so we discuss it in detail. But the overall risk profile is high now, there’s no doubt about that. We’ve got to be conscious of all of these threats. We do a lot of training internally, to prevent phishing attacks, prevent vulnerabilities, getting into the code.  

The proliferation of payment companies and fintechs, the speed in which they’re able to deliver, and the innovation, frankly, that we’re seeing in the marketplace — that’s something we’ve got to be ahead of. It’s the reason we’ve invested so much in our technology talents. We’ve hired from fintechs, cloud providers, as well as other banks. We’ve certainly tried to change our thinking in the bank, by adopting modern tools, moving to the cloud, and being able to attract equivalent talent. But some of these companies are unregulated and they’re able to operate at a speed that we can’t. So that’s something we’ve just got to be mindful of, how do we compete in that space?

Sponsored content

By Splunk

Financial services are already a top target for fraud, more than nearly any other industry. It’s a widespread problem: the majority of institutions have experienced at least one incident.

That’s why, for all the excitement around AI and machine learning (ML) right now, it’s understandable that some financial services leaders have concerns about what cybercriminals might do with open access to these new tools. It’s easy to imagine how bad actors could use AI to make their old schemes even more effective:

• Phishing schemes
  The industry has worked hard to train employees to see through phishing schemes and avoid falling into their traps. That email may look like it came from your boss, but you know she would never talk like that. AI could make the difference more difficult to discern. Attackers could use generative AI to capture the boss’s writing style — creating a much more convincing fake email. Or they might even use deepfakes that mimic her voice, making it much more likely employees will fall for the scheme.

• Business email compromise
  After working with the same vendors and managing the same invoices for years, it’s easy to slip up and pay a routine invoice without a second thought. But it’s not always so routine. Next-generation invoice fraudsters apply ML algorithms to identify patterns in communication to impersonate insiders at the companies they’re targeting. Then they send falsified invoices requesting payments that look real and are anything but.

• Account takeovers
  It’s not always a lowly cybercriminal manually entering passwords and guessing the answers to security questions. For many attackers, AI has made this once tedious process a whole lot easier. AI-powered botnets can deploy numerous brute force attempts, covering more ground in a shorter amount of time — and increasing the likelihood of success.

It’s already a nightmare if someone accesses your financial information — draining your account and killing your credit along the way. Now, imagine AI makes this easier for bad actors, and it starts to make sense why financial services organizations (and their customers) are worried.

Luckily, financial services defenders have access to the same tools — and new ways to use them. AI and ML tools can help analyze patterns across all sources of data, and even identify weak points that allow fraud to succeed in the first place. And these tools will make it easier for security analysts to operate. Instead of navigating an endless deluge of data, AI can help companies prioritize where to look to stop fraud before it can snowball into something worse.

Stopping fraud faster with AI and ML

In financial services, the fallout from fraud lingers long after the incident is resolved. Reputational damage can stick — customers want to do business with institutions they trust to protect their money and data. And reputational issues have big ripple effects, like drops in share prices, customer volume and market confidence. The total potential costs of fraud seem even more immense when you consider that the most disruptive incidents can cost as much as $50 million.

That’s why early detection is so important — the sooner you can see it, the sooner you can shut it down. Luckily, advancements in AI and ML could serve as an early warning system for fraud by spotting unusual patterns in the data. For example, they can identify unusual customer behavior trends — from what they’re buying to where they’re buying it, how they’re accessing their account to what time they’re doing so — and flag the issue for a closer look.

As they advance, AI and ML can even help outpace the fraud risks we haven’t seen yet. Too often, fraud and financial crime starts by exploiting a weakness that financial institutions don’t even realize they have. Modern tools that understand what “typical” looks like can help spot “atypical” before it becomes a greater issue.

Risk-based approaches that raise the right red flags

A single odd transaction might not mean fraud is underway, and no bank wants to unnecessarily shut down a legitimate one. Taking a risk-based approach can create risk scores based on suspicious activities with known risk indicators. If an incident reaches a predetermined score threshold, the system can alert the fraud prevention team to take a closer look.

So, yes, AI is going to raise the stakes for fraud prevention in the financial services industry. But odds are the right side will win.

AI is an accelerator for financial services visibility

AI and ML are getting all the buzz, but one underrated point is how they supercharge other equally important — albeit less hyped — advancements that help stop fraud. One big theme? Enhanced visibility and control of data. Cybercriminals are counting on being able to sneak in the back door. But the more visibility over your systems, the less likely they can act in stealth mode.

As leaders in the financial sector work to increase visibility across institutions, taking a data-driven and modern approach to cybersecurity is the way to go. Along with new advancements in AI and ML, solutions to consider include:

Breaking down data silos

When your systems don’t talk amongst themselves, it gets harder to see alarming data patterns. Connecting the dots between different data sets can bring patterns in consumer data to the surface.

Enabling collaboration across the institution

Many financial services institutions have several lines of business with different teams and tools working to detect fraud. While bringing the data together is a great start, giving people a view of the bigger picture helps everyone win (except attackers).

Find fraud faster with Splunk’s Machine Learning Toolkit

With Splunk, institutions can detect financial crime and fraud, and more accurately identify entities involved in suspicious behavior. With Splunk’s Machine Learning Toolkit, users can create a range of detections for fraud using algorithms and guided workflows. To prevent account takeover, the Machine Learning Toolkit can spot outliers in multi-dimensional datasets using clustering. Having analyzed historic user behavior, transaction activity and types of activity, the tool can identify atypical patterns in transactions and help lessen the blow from a takeover.

The banking and finance sector should develop a shared responsibility framework with their artificial intelligence partners to address fraud, scams and ransomware attacks arising from AI, Acting Comptroller of the Currency Michael Hsu said in June.

While speaking at the Financial Stability Oversight Council’s AI Conference, Hsu highlighted how AI’s use can threaten financial stability.

“From a financial stability perspective, AI holds promise and peril from its use as a tool and as a weapon,” Hsu said. “The controls and defenses needed to mitigate those risks vary depending on how AI is being used. At a high level, though, I believe having clear gates and a shared responsibility model for AI safety can help.”

Cloud computing’s shared responsibility model could be used as a blueprint, Hsu said.

“In the cloud computing context, the ‘shared responsibility model’ allocates operations, maintenance and security responsibilities to customers and cloud service providers depending on the service a customer selects,” he said. “A similar framework could be developed for AI.”

The newly established U.S. Artificial Intelligence Safety Institute, within the National Institute of Standards and Technology, can devise a shared responsibility framework from its consortium of more than 280 stakeholder organizations.

Hsu’s remarks come as Treasury Secretary Janet Yellen warned at the same conference that using AI in finance could lower transaction costs but carries “significant risks.” 

“Specific vulnerabilities may arise from the complexity and opacity of AI models, inadequate risk management frameworks to account for AI risks, and interconnections that emerge as many market participants rely on the same data and models,” Yellen said at the conference.

The Treasury Department is seeking public comments on the use of AI in the financial sector.

Hsu, meanwhile, used electronic trading as an example of technology that starts as a novelty and slowly becomes more trusted. AI follows a similar path, he said, because it was first used to produce inputs for human decision-making, then as a co-pilot to enhance human actions, and finally as an agent executing decisions on humans’ behalf.

However, banks adopting AI must now set up “clear and effective gates” between phases to ensure safety.

“Before opening a gate and pursuing the next phase of development, banks should ensure that proper controls are in place and accountability is established,” Hsu said.

Hsu noted the unexpected or unintended consequences of over-reliance on AI.

He cited Jake Moffat, who, in November 2023, took a chatbot’s help to find the bereavement rate on Air Canada’s website when his grandmother passed away. Moffat was booking a flight to Toronto. The chatbot suggested he book a flight and ask for a refund within 90 days, Hsu said. However, Air Canada prohibits retroactive refunds for bereavement flights. Moffat sued the airlines, but Air Canada argued it “cannot be liable for the information provided by a chatbot.” Moffat won the case, Hsu said.

With chatbots, companies can struggle to identify who is liable for what and how to fix it, Hsu said. The same questions arise when AI is used in credit underwriting, and consumers are denied credit cards based on an AI algorithm — a decision that is hard to explain, he said.

“With AI, it is easier to disclaim responsibility for bad outcomes than with any other technology in recent memory. The implications for trust are significant,” Hsu said. “Trust not only sits at the heart of banking, it is likely the limiting factor to AI adoption and use more generally.”

Hsu also noted the proliferation of deepfakes that have evolved from simple voice tricks, as well as more sophisticated deceptions used in big-ticket heists.

Though these incidents have been manageable, they could have a more significant financial impact as criminals become more advanced in using AI, Hsu said.

In December, Sens. Mark Warner, D-VA, and John Kennedy, R-LA, introduced legislation mandating that the FSOC coordinate regulatory efforts aimed at protecting markets from the potential disruptive impacts of deepfakes, trading algorithms and other AI tools that could destabilize the financial system. Yellen backed the legislation in February.

“An increase in AI-powered fraud could sow the seeds of distrust more broadly in payments and banking,” Hsu said. 

Artificial intelligence is poised to save financial advisers in Morgan Stanley’s wealth business hours each week, CEO Ted Pick said in June. 

Although the investment banking side is set to benefit from the proliferation of AI as well, the technology is likely to be “even more” of a game changer for the bank’s wealth business, Pick said.

One example he outlined: During meetings with wealth management clients, AI tools can transcribe conversations and categorize them by topic, saving the bank’s financial advisers 10 to 15 hours each week, Pick said at a Morgan Stanley conference. “That’s an enormous productivity quantum, and that’s not even getting to anticipating what might be the right conversation to have,” he said.

Pick took over as CEO in January, replacing James Gorman. As Pick detailed his expectations in June for how AI could change operations at the New York City-based bank’s investment banking and wealth management segments, he said the former is “going to be at the heart of a lot of this interesting development around AI.”

He envisions the broader AI opportunity creating “an enormous flywheel of activity” for Morgan Stanley’s investment banking business, pointing to the bank’s utilities clients who are buyers or suppliers of power and communications clients who are owners or lessors of towers, as well as AI companies themselves pursuing financing and initial public offerings.

Pick noted the importance of Morgan Stanley’s efforts to become a more integrated firm, because investment banking staff might be able to connect a large-scale tech company with those on the commodities side.

“That works at our place,” because there’s “a history of trust,” he said, adding, “That, over time, then starts bleeding into productivity.”

Pick emphasized that integration is a key element of his vision for Morgan Stanley.

“We’ve been unified for a long time,” he said. “You can’t just wake up one day and say ‘Let’s get along.’”

Morgan Stanley Co-President Dan Simkowitz will spearhead “the integrated investment bank continuity on steroids,” going after asset managers and asset owners with a holistic coverage mindset, Pick said.

Morgan Stanley’s size is a notable advantage when it comes to the cost benefits of technology use and improving efficiencies, Pick said. 

“You can afford to get ahead of cyber scenarios and bad actors and state-of-the-art surveilling,” he said. “And you can think about efficiencies associated with using AI to break down some of the classic infrastructure controls.” 

The bank’s CFO, Sharon Yeshaya, has begun working on the 2025 budget, Pick noted, and finance leaders are working with those on Morgan Stanley’s research team to consider how the efficiencies might work for the bank, he said.

“As long as there are high-quality eyes on top of the technology development, you can be in a cycle where you’re beginning to be able to process more [information] faster through the pipes, knowing that you’re going to have AI as something that will, over time, relieve some of the [operating expenses] pressure,” Pick said. 

Morgan Stanley in March named a firmwide AI chief, Jeff McMillan, who oversaw the development of AI @ Morgan Stanley Assistant, a virtual assistant for financial advisers and bankers powered by OpenAI.

JPMorgan Chase retained its leading position in artificial intelligence adoption across the banking sector this year, according to the Evident AI Index in October. The financial sector research firm ranked AI maturity among the top 50 banks based on talent development, public reporting and innovation and leadership metrics.

The bank has the largest AI workforce of the financial organizations Evident analyzed and employs more AI researchers than the next seven largest contenders combined. JPMorgan was also among the few financial companies to report return on investment on use cases, Evident said.

Capital One ranked second in AI adoption and made headway catching up to JPMorgan by doubling down on AI skills and development capabilities, according to Evident. The bank captured leading positions in density of AI talent relative to overall headcount and in innovation based on the number of patents filed. Capital One owned 38% of AI patents filed by 50 banks in the index.

As banks raced to develop generative AI use cases, JPMorgan balanced speed with prudence.

The bank took a top-down approach to adoption, adding Chief Data and Analytics Officer Teresa Heitsenrether to its technology leadership team in June 2023, and putting an AI assistant called LLM Suite in the hands of 140,000 employees in September.

Along the way, the company rolled out ChatCFO, a generative AI tool for finance teams and implemented prompt engineering training for new hires.

JPMorgan CEO Jamie Dimon prioritized data and infrastructure modernization to drive AI adoption in an April letter to shareholders, stating the bank planned to have 75% of data and 70% of applications migrated to the cloud by the end of the year.

“The characteristic of the top banks is that they started preparing five or six years ago with restructuring, hiring and putting in a research lab,” Alexandra Mousavizadeh, co-founder and co-CEO of Evident, told CIO Dive. “That’s what Jamie Dimon did.”

All four of the leading banks, JPMorgan, Capital One, Royal Bank of Canada and Wells Fargo, have dedicated AI research teams, Mousavizadeh said. “If you're going to scale your capabilities across multiple lines of business, you’re going to need a research hub,” she said.

JPMorgan, Capital One and Wells Fargo — the three largest U.S. players — employ 17.5% of banking’s AI talent pool, the analysis found. The AI workforce, which Mousavizadeh said encompasses roughly 240 roles and titles, grew 17% year over year across all 50 banks in the study.

The data-rich financial sector is uniquely positioned to lead the way in responsible adoption. “These are big legacy organizations that have a lot of data, a wide range of potential use cases and money to spend on compute power,” Mousavizadeh said.

As a highly regulated industry, banking has a vested interest in AI governance issues. JPMorgan, Capital One, Citi and several other large financial institutions are tackling compliance frameworks as part of the Fintech Open Source Foundation, an alliance that includes hyperscalers Amazon Web Services, Microsoft, Google Cloud and a number of enterprise technology vendors.

Banks lean on open source to attract talent, Mousavizadeh said. But erecting consistent guardrails and industry standards around model behavior remains paramount.

“Any bank using an AI-driven tool that comes out with something crazy is toast — they’re out of business,” Mousavizadeh said.

As the financial industry jumps to employ generative artificial intelligence, MoneyLion seeks to use artificial intelligence to match consumers searching for financial services with financial product partners. The fintech is integrating AI to increase efficiency while ensuring security and data privacy.

MoneyLion’s open-source technology enables it to present products from across the marketplace to consumers. 

The New York-based company aims to facilitate 25% of all financial product purchasing decisions through its platform within three years. If MoneyLion could achieve even a third of that vision, “this is going to be a very important business that's out there,” MoneyLion CEO Dee Choubey told Banking Dive. 

Some of the new products the fintech is building and offering its consumers are based on a prefilled identity graphs that analyze a consumer's preferences and anticipate other products they might need.

The goal would be to increase the MoneyLion network's conversion rate, which would benefit the supply side or the publishers embedding the platform into their websites, like CNBC, Fortune and Forbes. This would also benefit the demand side, which buys the leads, the executive said. 

The fintech's consumer business plans to have a generative AI-powered search engine by the end of summer.

Choubey said the service, available in a broad beta stage and using AI, is seeing great success.

Consumers can look at and “talk to their money” by asking questions about how much they spent on ride-share services on a particular day, month or year, or they can analyze activities related to their paycheck through modules and widgets by linking their own bank transaction data.

“Machine learning is a way of life at MoneyLion,” Choubey said.

“In a world of open banking, we want to be really the interface layer for MoneyLion,” he added.

Competing with business-to-business marketplaces, lead generators, publishers, neobanks, and payment companies for over a decade, MoneyLion is building out web services through the MoneyLion stack that is built on AI.

In February, MoneyLion announced its partnership with EY US to leverage the fintech’s embedded marketplace infrastructure by helping many smaller and mid-sized financial institutions offer more convenient capabilities to their customers.

Small regional banks may lack the monetary investment into their user experience like big banks such as JPMorgan Chase, but through partnerships like EY, MoneyLion’s platform helps clients reach at least halfway in providing these types of user experiences, according to Choubey.

‘Seismic impact’ of AI on fintechs

AI will transform the way consumers search for, identify and buy financial products by providing personalized recommendations that would otherwise require hours of manual search, Choubey noted. 

When consumer link their bank accounts to fintechs like MoneyLion and leverage those platforms’ network data, consumers with excess funds in checking accounts can be offered appropriate credit cards or recommendations on savings accounts. These “self-driving finance elements and modules”’ are likely to be implemented within the next 12 months, powered by AI, he added. 

Although still in early stages, integration of AI with personal devices and inter-app communication is expected to enable interaction between savings, checking and credit accounts, Choubey said. This could help customers avoid late fees, ensure timely payments, and consider alternative funding sources. 

“That is a pretty magical user experience and I think AI is going to take that to the next level,” Choubey said. “There are a lot of fintechs that are already doing it. If they're not going to do it, their business models will be impacted negatively.”

Challenges of using AI

Building consumer trust in AI remains a challenge, however. Choubey believes that can be overcome by increasing transparency around data usage, like giving read-only access to checking account data, clear communication on how recommendations are generated, and explicit consent from consumers for AI-driven actions. 

“If you forget your password to a social media app, the consumer is ticked off, but it's a mild inconvenience. But if we get any insights or activities wrong with their money, it's something that's going to lead to big consternation and big risks for everybody,” he said.

From a data security standpoint, MoneyLion has multiple layers of user-generated and consumer-elected data permissions. The fintech has also built numerous technologies to ensure context windows are only for financial transactions and limit the scope of products that use embedded gen AI.

However, Choubey thinks AI will augment human capabilities and that current roles, including call center agents and advisers, may evolve. Employees could be redeployed to more productive tasks or provide higher-level services, he said.

“I think it's overblown that jobs will be lost, but it's underestimated the growth impact that it will have,” Choubey said. 

The New York State Department of Financial Services issued new guidance in October advising financial firms to assess and address cybersecurity risks arising from the use of artificial intelligence.

The guidance, which imposes no new rules or regulations, highlights four AI-related risks that financial sector workers need to be aware of, including social engineering, cyberattacks, theft of nonpublic information, and increased vulnerabilities due to supply chain dependencies.

The regulator warned of the use of deepfakes – AI-generated videos, photos or audio recordings – which are sometimes created in an attempt to “convince employees to divulge sensitive information about themselves and their employers. When deepfakes result in the sharing of credentials, threat actors are able to gain access to Information Systems containing Nonpublic Information.”

“AI-driven social engineering attacks have led to employees taking unauthorized actions, such as wiring substantial amounts of funds to fraudulent accounts,” the regulator cautioned.

The regulator also warned that the prevalence of AI has increased the speed and scale of cyberattacks and given unsophisticated bad actors a “lower barrier to entry” to access sensitive NPI. Financial firms have a large volume of sensitive NPI, the NYDFS noted, making the sector particularly attractive to bad actors.

“I think it’s really about making sure there’s expertise in the institution, making sure they’re engaging with lots of stakeholders, so they understand the development of the technology,” NYDFS Superintendent Adrienne Harris said in an interview with the Wall Street Journal on the guidance’s release.

“It’s about making sure that you’ve got the right expertise in-house — or that you’re otherwise seeking it through external parties — to make sure your institution is equipped to deal with the risk presented,” she said.

Financial firms should have layers of cybersecurity controls that offer overlapping protections so that if one layer fails, there aren’t areas left unprotected, the NYDFS wrote in the guidance. Risk assessment and risk-based programs, policies, procedures and plans should be included in cybersecurity controls, as well as proper management of third-party vendors and access controls, including multi-factor authentication.

Cybersecurity training should also play a role in protecting NPI, educating all personnel on the risks of AI, NYDFS said in the guidance.

The guidance was issued in response to inquiries about how AI is changing cyber risk and how such issues can be mitigated. It comes a year after another regulator, Federal Reserve Vice Chair for Supervision Michael Barr, warned that the AI explosion had created a looming cybersecurity risk for the financial services industry.

Since then, a number of firms have fallen victim to cybersecurity breaches, though it’s unclear how many of them involved the use of AI.

“AI has improved the ability for businesses to enhance threat detection and incident response strategies, while concurrently creating new opportunities for cybercriminals to commit crimes at greater scale and speed,” Harris said in a prepared statement.

“New York will continue to ensure that as AI-enabled tools become more prolific, security standards remain rigorous to safeguard critical data, while allowing the flexibility needed to address diverse risk profiles in an ever-changing digital landscape,” she said.

The Biden administration is asking the Consumer Financial Protection Bureau to look at banks’ use of chatbots and to address customer service “doom loops” as part of a wider effort, unveiled in August, to prioritize customers’ time.

“Companies often deliberately design their business processes to be time-consuming or otherwise burdensome for consumers, in order to deter them from getting a rebate or refund they are due or canceling a subscription or membership they no longer want — all with the goal of maximizing profits,” the White House wrote in introducing what it dubbed the “Time Is Money” initiative. “Americans are tired of being played for suckers.”

Among the pillars of the initiative, the White House said the CFPB is planning to issue guidance meant “to crack down on ineffective and time-wasting chatbots used by banks and other financial institutions in lieu of customer service.”

“While chatbots can be useful for answering basic questions, they often have limited ability to solve more complex problems and disputes,” the White House said in its directive to the CFPB. “Instead, chatbots frequently provide inaccurate information and give the run-around to customers seeking a real person.”

Chatbots should “enhance customer service with speedy response times, not [help companies] shirk on basic responsibilities, such as receiving a refund,” the White House said.

The CFPB will identify when the use of automated chatbots or artificial intelligence voice recordings is unlawful, including when customers believe they are speaking with a person, the White House said.

“This is not about shaming corporations writ large,” a senior administration official told reporters Friday, according to CNBC. Rather, it’s “a new frontier of consumer protections.”

The initiative comes roughly a year after the Biden administration and the CFPB doubled down on efforts to root out unnecessary “junk fees” charged by banks for basic services.

“When people request basic information about their accounts, big banks cannot charge them massive fees or trap them in endless customer service loops,” CFPB Director Rohit Chopra told reporters last October.

Those loops, too, came back around in messaging August. The CFPB is set to float a rule requiring companies to let customers talk to a person by pressing a single button.

The White House cited a recent survey in which respondents reportedly said being forced to listen to long messages before being permitted to speak to a live representative was their top customer service complaint.

“Too often, customers seeking assistance from a real person are instead sent through a maze of menu options and automated recordings, wasting their time and failing to get the support they need,” the White House also said in August.

The CFPB isn’t the only agency being pressed into action. The Department of Health and Human Services is set to urge health plan providers to make it easier for customers to talk to live people, the White House said. The Federal Communications Commission, too, will launch an inquiry into how phone, broadband and cable companies can do the same. 

The Time is Money proposals will not require congressional approval, the senior administration official said. Any bill to boost consumer protections — this late in the Biden administration — would have a remote shot at passage because Republicans hold a majority in the House.

The White House’s initiative, though, comes as Vice President Kamala Harris, the Democratic nominee to succeed Biden in the White House, prepares to outline an economic policy.

Alternative lender Upstart has received a subpoena from the Securities and Exchange Commission regarding the fintech’s disclosures related to use of its artificial intelligence models and loans, the fintech revealed in its quarterly filing in May.

The San Mateo, California-based company said it received the subpoena Nov. 17, 2023; the regulator has asked for various documents and information. “We are cooperating with the SEC and are unable to predict the outcome of this matter,” Upstart said in the filing.

In its quarterly earnings report, Upstart noted that during the first quarter of 2024, its percentage of loans fully automated increased to 90% — up from 84% in the same period of 2023.

“For Upstart and our lending partners, it means there’s no human in the loop whatsoever to process and complete the loan application,” Upstart CEO Dave Girouard said during a call to discuss quarterly results, Bloomberg reported.

Upstart said it has nothing to add beyond what is included in the 10-Q in an email response to Banking Dive.

The SEC is one of multiple regulatory authorities scrutinizing financial institutions’ use of AI, including fraud prevention, credit underwriting and personalization of customer services.

The Consumer Financial Protection Bureau issued guidance last September on credit denials for lenders who use AI to make lending decisions. The Federal Trade Commission is investigating the partnerships and investments tech giants Microsoft, Alphabet and Amazon have with AI startups OpenAI and Anthropic, Bloomberg has reported.

The SEC subpoenaed Upstart about 17 months after the CFPB terminated Upstart’s no-action letter in June 2022. The fintech had urged the CFPB to terminate its no-action letter because it wanted to make time-sensitive changes like adding variables to its underwriting and pricing model — something that would interfere with the agency’s lengthy review process.

Upstart, which uses AI and alternative credit data such as a borrower’s education and job history as part of its formula to determine creditworthiness, became the first company to obtain a three-year no-action letter from the Consumer Financial Protection Bureau in 2017.

While sharing data with the CFPB in 2019, Upstart said its alternative data model helped it approve 27% more loans, with an annual percentage rate averaging 16% lower, during the first two years of its no-action letter period.

In July 2022, Upstart acknowledged in a press release that its loan marketplace had become “funding constrained” over a lack of demand from loan buyers. Additionally, challenging macroeconomic conditions continued to plague the fintech, which led to it laying off roughly 20% of its workforce, or 365 employees, last February. That followed November 2022 job cuts, when Upstart culled 140 employees.

Major fintech firms increased lobbying spending in the first half of 2024, with some hitting new spending records, according to a report from nonprofit OpenSecrets published in August.

Neobank Chime spent $1.05 million, nearly matching its total for 2022. PayPal spent more than $800,000, higher than the same time period during every prior year. And Block, which spent just under $1 million, hit its biggest six-month lobbying expenditure to date.

“That trend that new companies providing a new kind of service are getting involved in lobbying to limit the regulation of these new services from the very beginning is notable,” said Dan Auble, senior researcher at OpenSecrets. Companies lobbied on legislation covering taxes and payments, and on broader advocacy efforts on banking, cryptocurrency and artificial intelligence, OpenSecrets reported.

Many companies behind popular digital financial services — including peer-to-peer payments and earned wage access — spent more on lobbying in the first half of this year than any of their previous spending for the same time frame in previous year, OpenSecrets reported, citing filings the companies made to comply with the Lobbying Disclosure Act.

A major topic among PayPal’s efforts was increasing the $600 reporting threshold for business transactions on P2P payment apps and e-commerce platforms, established in the American Rescue Plan Act of 2021. The legislation significantly reduced the threshold — from $20,000 — for taxpayers receiving a Form 1099-K.

Meanwhile, some of Block’s lobbying initiatives focused on cryptocurrency and AI. Early Warning Services, the company behind P2P payments service Zelle, spent $180,000 for the first six months of this year, focusing on broader issues relating to payments systems, fraud and risk management.

Fintechs that offer earned wage access services also stepped up lobbying investments. Earnin — which spent $300,000 on federal lobbying during the first half of this year, more than the total amount it spent the prior year — focused on the Earned Wage Access Consumer Protection Act, a bill introduced in February by Rep. Bryan Steil, R-WI, that sets guidelines for the industry but exempts earned wage access payments from the Truth in Lending Act.

The Consumer Financial Protection Bureau in July proposed an interpretive rule under which some earned wage access payments would be subject to the Truth in Lending Act and other consumer loan protections. Neobank Chime also undertook lobbying efforts related to this topic, OpenSecrets reported.

“Three of [Chime’s] 23 lobbyists are officially working on specific legislation, the Earned Wage Access Consumer Protection Act, while the rest are working on categories similar to those of Early Warning Services,” the report said. “With about 30 legislative days left scheduled for the year in both chambers of Congress, the fate of a lot of the legislation being lobbied by the fintech sector may remain up in the air until the 119th Congress convenes.”

After Chime, the next-largest neobank to report lobbying spending was Varo Bank, which spent $20,000 a quarter since 2023 on broader issues related to digital banking, with efforts not tied to specific legislation.