Colony Bank uses gen AI to shave days off SBA loan approvals

• Key insight: Using generative AI helped Colony Bank adapt to new SBA rules more readily, according to an executive at the bank.
  
• Forward look: The bank plans to start approving larger loans on the platform in a few weeks.

When the Small Business Administration announced changes to SBA loan requirements in August, Darren Davis, president of Colony Bank's small-business specialty lending division, estimated the changes would add five days to the time needed to process each loan, on top of the typical 10-day loan approval process.
The Trump administration issued 416 pages of revised SBA rules that took effect Oct. 1. The new rules tighten lending criteria, increase documentation requirements and alter business acquisition standards.

Processing Content

Davis and other leaders at Georgia-based Colony Bank deployed generative AI-based business loan origination software from Lama AI and were able to reduce the lending process from 92 steps to 32 — a 67% reduction.

"Now if the customer's engaged with us and responsive, we're about seven business days," he said.

Colony now originates all of its SBA Express loans, along with certain other non-SBA commercial loan products, through Lama AI. 

In a broader trend, U.S. banks are seeing productivity gains from implementations of AI. In an American Banker survey conducted earlier this year, 41% of respondents said AI integration was the top driver of improved employee productivity over the last 12 months, followed by improved efficiency internally (34%), new tech and tools (25%), task automation (23%) and process simplification (17%).

In lending, banks are typically using AI in credit decisions to approve more loans. 

"If I'm using AI to say no, then that's questionable, especially under issues like fair lending," said Carey Ransom, managing director of BankTech Ventures. 

At PNC, generative AI is never used to say no on a credit decision, according to Ned Carroll, the bank's head of data and automation. "But I very well might use a generative AI system to say yes more often," he told American Banker in a recent interview.

At Colony Bank, generative AI performs data validation and analysis and makes a recommendation that's reviewed by a human.

The bank started down this path two years ago. "We had a small working capital program," Davis told American Banker. "We had two different tiers on it — up to $150,000 and up to $350,000 — and we were looking for a system that would allow people to get pre-qualified for one or the other before they went into a full application process, and so that was our initial thought on it. When we saw the robustness of Llama, we started going down a different path."

The different path was to use Lama as a full loan origination system, rather than just a front-end solution for initial assessments.

Davis' team used the software to remove 60 manual tasks from the loan process. 

"A lot of our information is document reading and verifying what the customer put in versus what's on a document," he said. "A lot of that went away because Lama does that for us. Lama's able to read documents and populate fields, so we know it's coming from the source and it's not being fat-fingered in. We're not having to double-check to make sure people entered things correctly."

The software compares data provided on loan applications against data from LexisNexis and a third-party business scoring provider to conduct know-your-customer and know-your-business data validations. 

Not having to double-check numbers saves an "immense amount of time for our team," Davis said. "It allows us to meet our goal of being able to spend more time on customer relationship building rather than monotonous tasks that really don't further a relationship."

The bank uses a pass-fail underwriting model. Lama "was able to take the information from different sources and understand if it passed that criteria or not," Davis said. "We still want a customer or a human to look at it."

Read more:

• As fintechs nab bank charters, what's happening to BaaS?
• Overheated data center knocked 23 credit unions offline
• Chris Britt still says banking is broken, but Chime can fix it
• Credit unions bulk up a business that banks long dominated

It took two months to roll out the software for the small working capital loan program. 

"We were under the gun because of the SBA rule changes," Davis said. 

Colony has now been using Lama for small working capital loans for about six months and plans to go live on it with larger SBA loans in a couple of weeks. 

The bank makes between 250 and 450 small-business loans a year. Davis said the bank can grow its working capital program with fewer people. One person who was processing originations has been moved over to the loan sales side already, he said.

"We're already saving on the gross cost" of SBA loans due to the new software, Davis said. "But once we see the returns down the road, it'll be even more compressive ROI than we have right now."

The bank is preparing to launch SBA Standard on Lama to originate larger business loans in the next couple of weeks.