Flex uses LLMs for 48 hour underwriting

This is really a claim about distribution as much as AI, because faster underwriting lets Flex approve and fund owner operated businesses before they drift back to Amex, a local bank, or a broker. In practice, the speedup comes from using LLMs to read messy bank statements, financials, and operating data that would normally sit in an analyst queue, then turning that into a credit decision fast enough to support a Net-60 card and working capital product for businesses in the $3M to $100M revenue band.

• Flex is lending off its own credit infrastructure, not just wrapping software around a bank product. It funds card float and Flex Capital through more than $300M of warehouse debt capacity, which means faster underwriting directly improves loan volume, pricing, and unit economics, but also makes the model more dependent on credit performance.
  1 sacra
• The target customer is the part of the market banks often handle poorly, profitable SMBs with uneven cash flow and lots of vendor spend. These companies send over PDFs, statements, and bookkeeping exports, not clean venture style metrics, so a model that can parse unstructured documents removes a real operational bottleneck.
  1 sacra 2 sacra
• The same document understanding layer also shows up in AP. Flex uses AI bill scanning, invoice extraction, approval routing, and accounting integrations with systems like QuickBooks and NetSuite, which means the underwriting engine and payables engine both get smarter from the same messy financial inputs moving through the platform.
  1 sacra 3 techcrunch 5 flex

The next step is that underwriting stops being a one time approval step and becomes a continuous risk engine tied to payments, invoices, and cash flow inside Flex. If Flex keeps compounding transaction history from card spend, bill pay, and global payments, it can price credit more precisely and make the lending product harder for software only rivals to copy.