Quick summary:
- 40% of the U.S. economy runs on small businesses, but banks allocate just 4% of lending to them, creating an access-to-capital crisis that AI-powered embedded finance is solving
- Pipe uses AI agents and real-time transaction data to approve loans up to $500,000 in minutes without credit checks or personal guarantees, embedded directly into platforms where businesses operate
- AI will make finance invisible by autonomously orchestrating bank accounts, credit cards, and cash flow predictions, removing the mental burden of financial management from business owners
Jake runs a well drilling business. He’s a veteran who survived a near-fatal accident and built a company where every employee owns shares. He’s also a skilled mathematician who managed his finances flawlessly but with crippling stress. He refused credit his entire life, didn’t own a credit card, and never considered borrowing money.
Then an embedded financing tool showed him his profitability in real time. He took a small loan, paid it back through his daily sales, and came back for more. The experience was so frictionless that it became a safety blanket. Now he’s replaced trucks, hired employees, and equipped his team with spending cards that don’t require credit checks. Jake discovered he was worthy of capital — and he’s not the only one.
On this episode of The Disruption Is Now, host Greg Matusky talks with Claurelle Rakipovic, Chief Product Officer and GM at Pipe, from the show floor at Money 20/20 about how AI agents are demolishing barriers to small business financing. The conversation covers everything from verification agents that compress 25 minutes of fraud checks into five minutes to the future of finance itself, where money management becomes so automated it disappears from conscious thought entirely.
Watch now:
Key takeaways:
Small businesses power 40% of the economy yet get just 4% of bank funding
The disconnect between small business economic impact and capital access is staggering. Banks allocate just 4% of their lending to the businesses that generate 40% of U.S. economic activity.
Rakipovic frames this as a worthiness problem. “Small businesses need to know that they are worthy of this capital, and that it is not whether they have a credit score or they have $50,000 to $200,000 in their account,” she explains. “It is the fact that they have the revenue, they have operations, that they run really well, and that at the end of the day, that is their collateral.”
The traditional lending system demands credit scores, personal guarantees, and mountains of paperwork. Pipe inverts this model entirely by using transaction data small businesses already generate. Four months of payment history can unlock loans between $1,000 and $500,000. No credit check required. No personal guarantee needed.
This matters because mental stress kills small businesses before bad decisions do. Restaurant owners operating on 3%-4% margins miss a single payment obligation because a fryer breaks. Well drilling companies face feast-or-famine cash flow. Medical spas can’t bridge the gap between service delivery and customer payment. AI-powered embedded finance can solve these structural problems.
Embedded finance meets businesses where they operate
Commerce moved to apps. Small businesses now run their entire operations through vertical-specific platforms that manage inventory, process payments, handle appointments, and track purchase orders. Banks still expect these businesses to leave their workflow, visit a branch or website, fill out applications, and wait days for decisions.
Pipe embedded financing directly into platforms like Uber Eats. Restaurant owners see loan offers right in their manager portal. Four clicks later, money appears in their bank account. The entire process takes minutes because Pipe’s underwriting runs on transaction data the platform already captures.
“You’re able to enrich the risk visibility you have of that business,” Rakipovic says. “And they have access to that [loan] instantly in a place where they’re doing their business every day.
The personalization extends beyond speed. Pipe knows when payroll hits. It can spot cash flow gaps four days before businesses run out of money and automatically extend bridge financing. Rakipovic’s team has prototypes running today that predict bank balances daily and intervene before NSF charges hit.
The embedded model also unlocks new products. Pipe offers cards that employees can use at Home Depot or Lowe’s, tracking receipts automatically without requiring credit checks. Businesses get working capital flexibility that previously required $50,000 to $200,000 in the bank plus pristine credit scores.
AI verification agents compress fraud checks from 25 minutes to five
Every loan requires Know Your Customer (KYC) compliance, fraud detection, and identity verification. These checks traditionally consumed 25 minutes of human mental load per application. Analysts cross-referenced postal codes and phone number variations, triangulated information across databases, and made judgment calls.
Pipe built an AI agent called Joe Verification to handle this work. Joe investigates 80 different flag types, conducts research across a wider spectrum of sources than humans typically check, and presents recommendations with stack-ranked priorities. What was once 25 minutes now takes just five.
Of course, the agent doesn’t replace human judgment. It compresses research and presents findings so humans can make faster, better-informed decisions. This matters because speed determines conversion in embedded finance. A restaurant owner with a broken fryer needs capital today, not next week. Five-minute verification enables same-day funding.
Finance will become invisible through autonomous orchestration
Banks know customers will overdraft seven days before it happens, but they don’t tell you. Rakipovic sees this changing completely as AI agents take over financial orchestration.
Today, business owners manually juggle bank accounts, credit cards, treasury yield products, and accounting systems while carrying intense stress about whether they’ll make payroll. Tomorrow, AI agents will move money between accounts, deploy credit cards when needed, check accounting systems for incoming purchase orders, and make autonomous decisions about when to extend credit.
The mental burden of cash flow management will shift from humans to machines that never sleep, never go on vacation, and operate 24/7.
Domain expertise becomes more valuable as AI handles execution
Rakipovic previously ran Amazon’s seller lending business worldwide, issuing up to $1 million in 24 hours using only Amazon’s transaction data. No additional information required. The program achieved unbelievably low loss rates and high returns on invested capital.
Her measure of success wasn’t profitability alone. Amazon tracked business growth. “I was able to drive 100% growth on average for [Amazon sellers],” she says. That growth created economic activity and small businesses across the globe that wouldn’t have existed otherwise.
The pattern she sees now is vertical specialization. As AI handles more execution, humans need deeper domain expertise to guide the machines effectively. “What’s happening with the world right now is you getting to a point where I’ll just call it a general theme of verticalization is people are getting deeper and deeper in their domain,” Rakipovic explains.
Her advice for her own children centers on this principle. She wants them to find big problems worth solving, develop enough curiosity to learn continuously, and build domain depth that lets them lead machines rather than being led by them.
Key moments:
- The gap between small business economic impact and bank lending (1:22)
- Why small businesses are worthy of capital without credit scores (2:51)
- How embedded finance meets businesses in their daily workflow (3:48)
- Real-time cash flow prediction that spots shortfalls four days early (6:10)
- Why restaurant owners need capital cards without credit checks (7:03)
- How Pipe uses AI to verify identity and flag fraud in five minutes (10:33)
- Jake’s story: from refusing credit to building business with embedded loans (13:18)
- The future where finance becomes invisible through AI orchestration (19:01)
- What Rakipovic learned scaling Amazon’s seller lending globally (22:18)
- Teaching AI about writing tone so non-experts can create quality content (35:47)
- Why AI should break traditional education models (37:10)
Q&A with Claurelle Rakipovic
Q: What’s the biggest gap in small business financing today?
A: 40% of the U.S. economy is run by small businesses, but only 4% of bank funding is provided to small businesses. There is a huge disconnect between what is powering our economy and what it is that we can do to help those businesses to continue to grow.
Q: How does embedded finance work differently than traditional lending?
A: Commerce has gravitated towards apps. Businesses are operating their whole operational life of the business in these platforms. What that data is, you’re able to enrich the risk visibility you have of that business and be able to assign them anything between $1,000 up to $500,000, and they have access to that instantly in a place where they’re doing their business every day.
Q: How do AI agents help with loan verification?
A: What used to take a human 25 minutes of web research, of really doing a lot of cross-referencing of information, agents can do that level of reasoning today. And not only that, they’re able to do the research on your behalf. We’re running that as the approach for the human then to cut their time from 25 to 5 minutes.
Q: What does the future of business finance look like?
A: Finance is not going to even feel like finance in the future. Your bank knows that you’re going to have an NSF charge seven days before you do. I have so much information today that I can predict your bank balance daily. I can look at your bank account. I can move money between your bank accounts if you let me. That is going to become autonomous.
Q: What should the next generation learn about AI?
A: What I expect from them is that they can find big problems that they themselves can start to work on. Use machine learning, use agents, code themselves, and that level of agency is possible today with a good degree of curiosity. You don’t actually need a high level of skill. What you need is a brain that can find the problems that are worth solving.
Q: Why is domain expertise becoming more important with AI?
A: What’s going to be expected of the human is that they hold some level of domain expertise. You must still study something and learn it just for the fact of going deep and not being at the surface and learning how to critically reason. Whether it’s a franchise or whatever platform it is, people are getting deeper and deeper in their domain, and in a sense, you need to own that and lead the machine.
