A $5 billion-asset credit union and one of the country's largest banks are experimenting with approaches to .

and in Lake Jackson, Texas, teased their plans in a pair of panels at this week. For smaller institutions, digital banks and fintechs, the quest for digital lending "is all about scaling the entire customer base," said Barath Narayanan, the head of digital engineering company Persistent Systems' global banking, financial services and insurance unit, on a panel.

For large institutions engaging in commercial lending, "It is all about process digitization right now," he said.

Whether lending to retail or commercial customers, speed is of the essence.

Large banks with commercial clients find that "there is a consistent focus on turning around faster term sheets," said Rich Longo, senior advisor at McKinsey, at the conference. These institutions are trying to reduce the time it takes to underwrite commercial clients from two to three weeks down to a few hours, he said. This is easier as tools to automate aspects of lending become more widespread, including data sources that are now digitized, such as IRS tax transcripts.

McKinsey research shows smaller and mid-cap businesses in particular will take the first offer they get as long as they like the conditions, rather than wait to see what rates they get from other banks, because they need the funding to operate their businesses.

"We're seeing that over 70% of the time, there is not as much rate shopping because they expect the banks to be within a certain parameter," said Longo.

The nimbleness with which a financial institution operates in the digital lending space is affected by the size and complexity of the company.

"Citi is a big ship to turn," said Tiffany Patrick, senior vice president of AML payments and innovations and innovations at Citi. Large institutions such as Citi that have acquired fintechs and other startups must meld disparate technology stacks that "are not necessarily meant to talk to one another," she said.

The bank is currently identifying ways to automate lending, from document submission to validation.

Meanwhile, TDECU has aspirations of a co-branded credit card for a convenience store chain that will prioritize efficiency for the applicant. It will largely target customers within the state of Texas.

"One of the things that we want to differentiate ourselves in the market is, can someone scan a QR code and in less than a couple of minutes be onboarded?" said Ashish Chopra, chief information officer and chief technology officer at TDECU.

Moreover, if the customer is already going through compliance checks during the credit card onboarding process, Chopra hopes to make subsequent applications for other credit union products easier by using the information already captured.

TDECU is assessing multiple vendors as it develops the card, and hopes to launch in the fourth quarter of 2024.

The limits of artificial intelligence is also an open question.

"It's POC [proof of concept] heaven out there," said Longo about using generative AI in lending, while emphasizing that less buzzy technologies such as robotic process automation are still extremely useful.

Large institutions run into problems when a large portion of their data is not normalized and there is little consistency as to which of their systems are in the cloud or have API connectivity.

Patrick of Citi said on her panel that normalizing data and ensuring its accuracy is key before the bank can use it effectively. There is also the challenge of and ensuring it is , which holds up use cases from officially launching.

"From someone who consistently has to speak to regulators during exams, I cannot say, 'we will just send it to that engine and it does its thing,'" said Patrick.

In her view, her entire team needs to understand the basic mechanics of what is being done with AI, not just the person explaining its usage to a regulator.

"This is unproven to the regulators so if you make a mistake, for the next few years that will end any of your automation projects," said Longo. "On the IT examinations, you know exactly where they're going to go first."