Inside LoanPro's AI Bet: Guardrails Before Growth

Farmington, Utah — July 15, 2026

As lenders race to bolt AI agents onto servicing and collections, LoanPro's approach is less about the chatbot and more about the guardrail. Chief Product Officer Colin Terry describes the Farmington-based fintech's new MCP-based AI Gateway as a system built on layers of compliance guardrails, regulatory boundaries an AI agent physically cannot act outside of, no matter what it has been asked to do.

It's a telling emphasis for a company that has spent the last several years expanding well beyond its original footprint as a loan servicing platform. LoanPro now supports card products, works directly with banks and credit unions rather than only their fintech partners, and is preparing international expansion into the UK, EU, and potentially Hong Kong. But in a recent conversation, Terry was just as focused on where the company draws the line as on where it's growing.

The overhead projector problem

Terry's explanation of how LoanPro thinks about AI and compliance leans on an unexpected metaphor: the overhead projector.

"You had an overhead projector in school, you could see that transparency on the wall," Terry shared with TechBuzz. "You lay a transparency down with a circle on it, then you put another transparency on top with a triangle... it's this sort of Venn diagram of where you're allowed to operate."

Each transparency represents a different layer of regulatory obligation, such as the CFPB, the OCC, the FDIC, card network rules, sponsor bank requirements, internal compliance policy. Stack them all together, Terry said, and what's left is the narrow, overlapping space where a lender is actually allowed to act. LoanPro's compliance guardrail system, he said, encodes that overlap directly into the API layer, so an AI agent attempting to execute a servicing action, such as adjusting an interest rate, offering a hardship plan, proposing a settlement, gets stopped the moment it tries to operate outside that space.

"Even if they think they're doing something great for the client," Terry said, the system won't let an agent "color outside the lines."

That framing matters because it reframes what LoanPro's MCP-based AI Gateway is actually for. Rather than layering a chatbot on top of existing workflows, Terry described a system meant to let AI agents take real action — running hardship plans, offering settlements, executing servicing decisions that would normally require a human agent — with the compliance boundaries built into the infrastructure itself rather than left to the judgment of the model.

Staying in the credit lane

That same instinct for boundaries showed up when Terry was asked directly where LoanPro won't expand. The company has grown aggressively since raising $100 million in 2021, moving from a pure servicing tool into card products, direct originations work with banks and credit unions, and now agentic AI tooling. But Terry drew one clear line: LoanPro will not build deposit cores. (Deposit cores are the underlying software infrastructure banks and credit unions use to manage deposit accounts — handling balances, transactions, and the ledger of record for money on deposit, as distinct from lending or credit products.)

"What I'm saying is off limits is deposit cores, and that's partly because we feel they're better partners than competitors at this point in time," Terry declared.

It's a deliberate choice, not a resource constraint. LoanPro has built relationships with core providers including Nymbus and Synctera, and is deepening ties with legacy providers like Jack Henry and FIS. Terry described these relationships as increasingly collaborative rather than adversarial. Asked why the company doesn't compete more directly with the deposit-side infrastructure it integrates with, Terry offered a blunter version of the same answer: "Got to be careful when you throw that stone at Goliath's forehead."

Instead, Terry said LoanPro is staying "squarely in the lane of credit products, both installment line of credit and credit card," pointing to shifts in how younger consumers build relationships with lenders. Buy-now-pay-later products from companies like Affirm and Klarna, he argued, have shown that credit products, not deposit accounts, are increasingly becoming the entry point of the borrower relationship. That shift, in his telling, is exactly why LoanPro sees more room to grow within its existing lane rather than a need to expand outside it.

Built for builders, not just operators

LoanPro's expansion hasn't been limited to product lines. It has also visible in the growing range of companies running on the platform, from large fintechs to banks building card programs from scratch. Terry pointed to Intuit as the clearest example of what that growth can look like at scale.

Intuit first came to LoanPro to migrate a single loan portfolio, mainly to regain full control over its own data rather than viewing it filtered through a third-party servicer. Once live on the platform, Intuit's innovation team ran a small pilot — an installment loan program offered to a handful of preferred customers — before scaling that program to 100 customers, then 1,000. From there, Intuit built the product directly into QuickBooks Capital, and eventually into its tax platform as well. Today, Terry said, that tax-season lending product originates anywhere from 10 to a peak of 50,000 loans a day, depending on the volume of tax filings on a given day.

Supporting that range of customers, from large, highly technical engineering teams to smaller, less tech-forward lenders, is itself a design challenge, Terry stated. LoanPro builds around three categories of user: the frontline human agents who operate inside the platform's UI, sophisticated technical teams who treat LoanPro as backend infrastructure to build on top of, and lenders in between who need to see AI-powered workflows in action before they'll adopt them. The AI Gateway, in Terry's framing, is meant to serve as a bridge for that last group. It gives them a working example of what's possible before asking them to build on the API themselves.

What's next

LoanPro's international push is similarly incremental rather than a flag-planting exercise, Terry said, driven largely by existing partners with international presence asking LoanPro to follow them into new markets. The company already operates in Australia, Canada, and has customers in Mexico, with UK, EU, and potentially Hong Kong markets under consideration for 2027.

Terry also confirmed the company is expanding relationships with payment infrastructure partners as its card business grows, though he was careful to characterize those relationships as partnerships rather than dependencies. For now, LoanPro's stated strategy is narrower than its growth might suggest: stay inside the credit lane, build the compliance guardrails deep enough that AI agents can be trusted to operate inside them, and let expansion follow wherever partners and client demand lead next.

LoanPro's Salt Flats Summit

Terry's conversation with TechBuzz also touched on a company tradition that's become central to LoanPro's culture: the company's annual Salt Flats Summit, an invite-only executive forum held out at the Bonneville Salt Flats. Last year's event drew more than 200 industry leaders, and organizers say this year's gathering is expanding further.

The format mixes standard conference programming with something considerably less conventional. Attendees spend part of the summit in hotel sessions before boarding buses for a two-hour drive out to the salt flats themselves, where LoanPro sets up a full tent city, a stage for product announcements, helicopter and monster truck rides, and a group fire walk.

"It's a little bit like some weird combination of a product kind of tech conference, Breaking Bad, and a little bit of Burning Man flavor," Terry said, describing the event's deliberate push to get executives — including LoanPro's own leadership — outside their comfort zone.

Details of the event, taking place this year on September 16 and 17, are available at saltflatssummit.com.