When talking about machine learning algorithms, Amazon definitely comes to mind.

The e-commerce giant has (arguably) the most sophisticated ML toolset around that is uses for drone deliveries, online tailors, or its AWS offering. Turns out, Amazon’s lesser known operation – small business lending – is also powered by machine learning, Bank Innovation has learned.

Since the launch of its SME lending program in 2012, the company has issued more than $3 billion in total loans, with more than $1 billion issued in the past 12 months, Amazon announced today. In total, more than 20,000 “invited” businesses received those loans.

So, how does one get “invited” to this exclusive SME club? According to a company spokesman:

We have a machine learning model. Amazon loans are offered on an invite-only basis to sellers on Amazon who qualify based on various criteria, including account tenure, and meeting the highest levels of customer experience.

Amazon did not disclose any other details around the algorithm (but I will think twice before leaving a bad review for sellers).

Through the program, businesses can apply for loans from $1,000 to $750,000, via a “convenient and easy” process, the spokesman said. “Invited sellers receive a pre-populated invite through their Amazon Seller account. Sellers give us a few pieces of information and the loan is extended same-day or next day in most cases,” he said.

Amazon Lending was created to help SMEs “expand inventory and operations at a critical period of their growth. We understand that a small loan can go a long way,” Peeyush Nahar, VP for Amazon Marketplace, said in a statement. Around half of the Amazon borrowers take out a loan the second time, the company said.

The program is currently available in the U.S., U.K., and Japan; the company is “evaluating future expansion,” the spokesman added.