Consistent underwriting at scale — and decisions cut from 45 minutes to 15

The Challenge

Scaling a lending operation without scaling the risk

Adam Mitchell has run this business for over 12 years — a former executive at a publicly traded bank and CEO of a regional bank. He knew from the inside that the only way to scale was a front-end AI platform: one that pulls in real, live data and pushes it back out, so every credit decision rests on the same foundation.

The harder problem wasn't speed — it was consistency. At volume, the way one underwriter reads a bank statement, a cash-flow trend, or a Secretary of State finding has to match the next. Getting that right on the front end was non-negotiable.

“The only way to manage volume is to make sure the process and workflows are consistent between one person and the other. The way we look at bank statements, cash flow, the Secretary of State findings has to be the same for every deal.”

The Solution

One consistent front end, wired into the loan platform

Kaaj sits at the very front of CSP's production credit process. It takes the credit package, runs the analysis, then populates their loan-platform system through an open API — the two talk to each other automatically. The result is a front end that's both fast and uniform: the same rigor on deal number one and deal number ten thousand.

Adam's favorite piece is the bank statement analysis — inflows, outflows, debt service coverage, and a fraud check that together reveal the true cash flow of a business. And the platform is easy enough that onboarding a new hire doesn't take days of training sessions.

“It cut our decision-making time in half. If we're faster with our decisions, the deal gets booked quicker and we originate more. It's the fastest I've ever seen the business work in over 12 years.”

The Results

Faster, cleaner, and built around the business

A complete package that used to take 45 minutes to underwrite now finishes in about 15 — and just as importantly, it finishes the same way every time. For a lender processing thousands of loans, that consistency is a risk control: a 35-to-40-page credit package gets parsed for exactly the data points that matter, without the mistakes a human eventually makes at volume.

Having spent a career with the big IT vendors, Adam expected to be locked inside someone else's rigid product. Instead, CSP has customized Kaaj from day one as the business has grown — the opposite of being stuck in someone else's box.

Next on the roadmap: a front-end portal, now in beta, to replace the email tree — and a path toward a single vendor across origination and the loan platform.

“Our credit package might be 35, 40 pages, and Kaaj picks out the particular data points we need versus a human doing it.”