May 2

By analyzing central bank messaging for its level of dovishness or hawkishness and finding patterns in how it has historically affected markets, JPMorgan’s tool shows how AI models can give banks an edge in trading. 

JPMorgan debuted a ChatGPT-powered model for detecting the tenor of central bank messaging that can help interpret current signals and, ultimately, predict upcoming market shifts. 

It analyzed 25 years of Fed transcripts to score them based on how hawkish or dovish the statements were and then compared those scores to historical market moves. 

“Plotting the index against a range of asset performances, the economists found that the AI tool can be useful in potentially predicting changes in policy — and give off tradeable signals,” according to Bloomberg. “For instance, they discovered that when the model shows a rise in hawkishness among Fed speakers between meetings, the next policy statement has gotten more hawkish, and yields on one-year government bonds advanced.” 

The Fed is expected to raise its benchmark interest rate again this week, and JPMorgan economists said that preliminary applications of the model ahead of the Fed’s meetings are “encouraging.”  

The tool – which produces what JPMorgan refers to as its Hawk-Dove Score – will be applied to data from more than 30 central banks around the world in the coming months, the bank told Bloomberg. It also provides another example of how FinServs can experiment with OpenAI’s buzzy ChatGPT technology

Find more of Insights Distilled’s coverage on how generative AI is shaping financial services here.

New research provides concrete evidence that generative AI can drive increased productivity for customer service agents, especially new recruits, by providing live scripts and linking to technical information. 

There’s new evidence on how generative AI tools can supercharge customer service agents.  

A study from Stanford and MIT researchers found that when ~5,000 customer service agents at an unnamed Fortune 500 company received real-time suggestions from a chatbot for how to respond to customers, they resolved 14% more issues per hour.   

The AI would monitor customer chats and provide wording suggestions for how agents could respond, as well as links to technical information that could help them troubleshoot problems.  

The chatbot was particularly useful for new, less-experienced agents: It essentially acted as a way for recent recruits to automatically receive advice without more experienced agents having to personally pass it on. The study also found that the tool improved customer sentiment and reduced employee turnover overall.

While anecdotes and predictions have pointed to generative AI’s ability to increase productivity, this study is some of the first empirical evidence showing its impacts in the real world. Notably, the AI tools weren’t replacing customer service agents, but augmenting them by helping them do their jobs more efficiently.   

Fairness in lending is historically fraught, and experts worry that opaque AI models could inadvertently make discrimination worse. To avoid that outcome, researchers argue that combining new data with fairness constraints and better models could both advance equity and protect lenders.

A combination of sophisticated AI models and historically off-limits data could create more equity in lending. 

Researchers from the Federal Reserve Bank of Philadelphia published a paper that argues that sophisticated AI-based credit models should use data that’s currently barred due to the Equal Credit Opportunity Act – like applicants’ location information – to make lending fairer.  

Their basic rationale is that removing sensitive attributes like location hasn’t historically solved the problem of systemic credit gaps. Instead, those attributes should be deliberately integrated into modeling in a way that actively corrects their effects, the researchers say.   

For example, lowering credit thresholds for people in lower-income areas could be a “fairness constraint” that gets integrated into lenders’ modeling.  

To integrate these fairness constraints and drive more equitable results while also mitigating losses, lenders would need to use more complex AI-powered models that can predict risk better than older methods, researchers say. 

“It is crucial to combine the introduction of fairness constraints with better machine learning models,” Vitaly Meursault, one of the researchers, told American Banker. “That will allow lenders to predict default better and compensate the costs of the introduction of fairness constraints, while at the same time reducing credit access gaps to creditworthy consumers.” 

Ultimately, the paper posits that “if the trade-off is managed appropriately, incentives can change in a way that both fairness and profits can improve over time.” In other words, there is a path forward for integrating artificial intelligence into lending that makes the whole system fairer without increasing risk. 

Read more about the research and its potential implications at American Banker. 

Despite regulatory uncertainty and recent turmoil in the crypto markets, Visa and Mastercard are doubling down on efforts to prepare for blockchains to become mainstream payments infrastructure.

Visa and Mastercard have both indicated their commitment to crypto in recent days through job descriptions and product updates. Their investments are a form of futureproofing: If blockchain technology does ultimately transform payments, the stakes are high for them to be on the forefront of that transition.  

For example, Visa posted new job openings for software engineers to aid its “ambitious crypto product roadmap” and help drive “mainstream adoption of the public blockchain networks and stablecoin payments.”  

Insights Distilled has previously reported on the firm’s “DeFi mullet” strategy, which revolves on the assumption that the blockchain will eventually become a standard backend technology that users don’t need to understand to use. In other words, fintech in the front, DeFi in the back.   

Visa’s head of crypto also called out on Twitter that the firm is looking for candidates who have used AI-powered engineering tools like GitHub Copilot to write and debug smart contracts.   

Mastercard, meanwhile, is teaming up with Web3 players for an on-chain identity verification framework. The firm’s so-called Crypto Credential initiative aims to boost trust for blockchain transactions.  

“Providing access to crypto in a safe way is also part of our value proposition and we’re continuing to do that,” an executive told Reuters, adding that Mastercard currently has dozens of partners around the world and is continuing to expand its reach.  

Identity verification has long been a cost center for banks, but IDPartner wants to flip the script. It’s building a network that would let businesses prompt customers to identify themselves by logging in with their bank IDs.

Banks invest huge sums in building safe, accurate know-your-customer (KYC) verification systems – and they should be able to offer those efforts as a service, according to a group of former Santander executives.  

IDPartner is creating a network that lets businesses prompt people to log into their sites using their banking credentials. The system will give firms confidence that their customers are who they say they are, since they’ve been vetted and authenticated by their banks.  

Meanwhile, the system gives consumers an easy, secure way to confirm their identity – through a button that “works like a social login from Google or Facebook, except it connects to the user’s bank,” CEO Rod Boothby tells Insights Distilled.  

It ultimately lets banks turn their investment in KYC into a source of revenue (each time a business gets a successful ID verification, it would pay a small fee), while also building deeper relationships with customers. 

“By giving banks the tools to manage digital identity as a new asset class and by aligning marketplace incentives to drive adoption by companies and users, we can transform the most frustrating identity journeys into an experience as easy and familiar as social login,” according to Boothby.  

The company just raised a $3.1 million seed round of funding and is modeled after Norway’s BankID ecosystem, which has 99% adoption in Nordic countries.