There’s a saying in mortgage lending that Robert S. Stewart Jr. shared early in our conversation, and it stuck with me for days afterward: the real underwriting doesn’t happen until 72 hours before closing.
Think about what that means for the person on the other side of the table. They’ve filled out the application. They’ve uploaded bank statements, tax returns, payoff letters and ID. They’ve paid for an appraisal. Earnest money is on the line. They’ve waited weeks. And only then does someone look closely enough to tell them whether the deal was ever going to work.
Robert has been that borrower. He described signing up for terms, getting to the end of the process, and learning the day before closing that the lender, the rate and the leverage had all changed — and that he needed to bring tens of thousands of dollars more to the table than he’d planned a week earlier.
So he built a company to move that moment from the end of the process to the very beginning.
I’ve spent most of my career inside financial institutions, from Merrill Lynch and Citibank to Credit Suisse and TD Bank. I know why banks still run decades-old systems for underwriting. The process is complex, heavily regulated, and when something works, nobody wants to touch it. That’s exactly why Robert’s story is worth studying. He didn’t just add AI to an old process. He asked where the real decision was happening, and then rebuilt the system around making that decision earlier.
Start with the hill you need to take
Before founding CR Equity AI, Robert was selling real estate and originating loans. Before that, he was an officer in the U.S. Air Force. That second chapter shaped how he built the company.
In the military, he told me, they used something called objective-based planning. You identify the hill you need to take and why, and then you work backward from there.
When Robert set out to build his platform, the objective was specific: an actionable term sheet in under five minutes, with roughly 95% confidence that the loan would close based on the information provided. Questions like who funds the loans, which asset classes to cover and which markets to serve were, in his words, things he could figure out later. The objective came first.
This is the discipline I see missing in most AI initiatives. Companies start with the technology — a model, a vendor, a pilot — and go looking for a problem. Robert started with a borrower’s pain point and a measurable outcome, and only then decided what to build. He reviewed the open-source underwriting code that was available, decided none of it met his standard, and built a fully custom engine.
“No” is the easiest answer
The most important insight from this episode isn’t technical. It’s cultural.
Robert pointed out that it’s easier for an underwriter to say no, because nobody gets fired for declining a file. That incentive quietly shapes an entire industry. Files sit in queues. A decline comes with little explanation. The borrower is left guessing.
CR Equity AI takes the opposite stance. When a deal doesn’t fit, the platform doesn’t just reject it. It returns the conditions and a path to cure them — templates, follow-up questions, and a plan to address the problem. Robert was clear that this doesn’t mean lowering standards. His loans still require background checks, minimum credit scores, experience verification and liquidity verification. The difference is that the system is designed to find the route to yes rather than the reason for no.
Every organization has its own version of the underwriter’s “no.” It’s the procurement process that defaults to rejecting new vendors. It’s the compliance review that’s rewarded for slowing things down. It’s the approval chain where nobody is penalized for waiting. AI won’t fix those dynamics on its own. But it can make the cost of a slow “no” visible, and it can make a fast, well-reasoned “yes” much easier to deliver.
Continuous underwriting, not a final exam
Robert calls his approach Kaizen underwriting — continual discovery and vetting from the beginning of the process all the way to the end.
In a typical process, a borrower submits high-level information, gets preliminary terms, then comes back later with a stack of documents, and the deep review happens at the end. On Robert’s platform, the borrower answers a short application and shares their documents or account access up front. The system builds a profile and measures it against the underwriting criteria immediately. Then it asks for the stipulations needed to support what it found, instead of discovering surprises in the final days.
He also made a point I think every operations leader should hear. In a traditional workflow, if the analyst who owns your file is out sick, on leave or overloaded, your file waits. His underwriting agents run around the clock. He described checking his phone on a weekend and finding a set of deals already screened and ready for his approval.
That’s the real promise of AI in operations. Not replacing judgment, but making sure work doesn’t stall while waiting for a person to pick it up.
Transparency as a competitive advantage
Two decisions Robert made run against nearly everything I’ve seen in financial services.
First, CR Equity AI is deliberately not a loan marketplace. Robert explained that many marketplaces send a borrower’s application out to dozens of lenders, each running its own credit check, which can damage a borrower’s score and lead to a flood of calls. His platform gives the borrower one point of contact and one set of underwriting criteria, and matches the deal against partners’ requirements behind the scenes.
Second, he publishes his underwriting criteria, fee policy and process openly. Most lenders guard those details closely. Robert’s view is simple: if you show people exactly how to succeed, they come prepared, the deals get better, and you spend less time on meetings with people who don’t actually have a deal.
We tend to think of a competitive moat as something hidden — proprietary data, a secret process. Robert’s moat is partly the opposite. The trust he builds by being transparent is something competitors can’t easily copy, because it would require them to change how they operate.
Compliance that moves at the speed of the market
Financial services is one of the most regulated industries in the world, and that’s often the reason it adopts new technology slowly. By the time a new tool clears every review, it can already be dated.
Robert’s answer is to build compliance in from day one and keep it current continuously. He described two practices.
The first is a two-gate model. Every deal must satisfy both CR Equity AI’s internal criteria and the eligibility rules of the eventual note buyer — the institution that will purchase the loan after closing. If the two sets of requirements don’t overlap, the deal doesn’t happen. He contrasted this with lenders who close a loan first and worry about finding a buyer afterward.
The second is a set of AI agents that monitor the market and flag changes for leadership in real time. When major lenders tighten their criteria or conditions shift, Robert gets an alert and can decide quickly whether his own underwriting needs to adjust, rather than waiting for the next committee meeting.
For any leader in a regulated industry, this is the reframe worth taking away. Compliance doesn’t have to be the brake on innovation. Designed well, it becomes part of the product.
Slow down to speed up
When I asked Robert about expanding into new areas, he told me he sees his technology applying in ten to fifteen other places. Then he explained why he isn’t rushing into any of them.
He described attending the Inc. 5000 event and meeting founders with far more revenue than he had at the time, whose products didn’t live up to their pitches when he tried them over lunch. His takeaway was to slow down to speed up. He’d rather be the Ford pickup truck of the tech industry — not the company with a new model every season, but the one whose product lasts.
He has acted on that. When a broker portal started pulling focus from his core business, he shut it down and rebuilt it with AI agents for compliance and underwriting. The portal now has far fewer brokers, but the deal flow and quality are consistent. He also turned down an eight-figure acquisition offer because, as he put it, his work there isn’t done.
This is the same principle at the heart of my own framework: before you branch out, you have to own your nest completely. The branches only strengthen the core if the core is strong.
Pay once, cry once
The people side of Robert’s business may be the most unusual part of all. He lets his employees tell him what they want to earn, and he rarely pushes back. One of his top performers asked for double his salary — and came prepared with his case.
Robert’s reasoning comes from a saying he grew up with in Alabama: pay once, cry once. He also described a military version, which he called the 1-3-8 rule. Getting something right costs a dollar. Changing course midway costs three. Rework costs eight. Applied to talent, the math is clear: paying great people what they’re worth up front is far cheaper than losing them.
He adds incentives for team members and partners who find and document ways to improve the platform, and he has built much of his team by growing people internally, including interns who have gone on to top academic programs.
The question for your organization
Robert closed by saying that, despite how far the company has come, the core idea is simple: give people transparency and access to capital in real time. It shouldn’t be revolutionary. But it is, because so few have done it.
That’s the challenge I’d leave with any leader thinking about AI. Look for the place in your organization where the real decision happens too late — where work sits in a queue, where “no” is the default, where people only find out the truth at the end. Then ask what it would take to move that decision to day one.
🎧 Listen to the full conversation with Robert S. Stewart Jr.:
Apple Podcasts: https://podcasts.apple.com/us/podcast/what-comes-next-with-arun/id6788060249
Spotify: https://open.spotify.com/show/033KRXb9RHJJIDSoE2KjYF
YouTube: https://youtu.be/OlKeJfiyIbc