What Your Best Underwriter Knows (And How to Make Sure the Whole Team Knows It)

Every community bank and credit union lending team has one. One person who, if they left, the whole house of cards would topple.

This person is the underwriter who has been there for twenty-two years and knows which borrower profiles pass without a second look. The indirect auto funder whose Saturday shift clears twice the volume of anyone else’s weekday. The processor who catches the title mismatch everyone else misses, every single time.

They are the engine of the lending desk. They are also, in the language of risk committees, your single biggest piece of key-person risk.

When your most experienced operator retires or takes a job elsewhere, the whole team’s throughput and accuracy drop with them. The new hire takes months to reach competence and years to reach their predecessor’s level. The institution absorbs the cost in cycles per file, condition pile-up, dealer complaints, and QC audit findings. The members also absorb it in longer processing times, uncertainty, and stress while waiting for an approval they desperately need.

The industry acknowledges the problem - we all know we have it. Until recently, we haven’t been good at solving it.

AI as a succession plan

The institutions most prepared to solve this are looking to agentic AI. They are leveraging it not just as a productivity tool, which is the most common and talked-about use case, but as a knowledge-capture tool. The judgment of your most experienced operators gets written into the workflows the agents follow. Every subsequent file benefits from that judgment, applied the same way every time, by every team member, in the system of record.

This use case shows up across the entire lending workflow. In indirect auto, it is the funder’s eye for the dealer whose contract packets routinely show a missing back-end product form. In HELOC, it is the team lead’s read on a lien position mismatch everyone else would miss on a second review. In mortgage, it is the senior underwriter’s acceptable-variance call on income calculation for a borrower with complex pay. In every lending workflow, there is a layer of judgment that lives in a few people’s heads and makes the whole team’s output better. Until it is captured, that judgment is also non-portable.

What capturing it actually looks like

At the customers we work with, codification happens in a defined partnership between the institution’s most experienced operators and the Saris team. The senior operator walks us through the real decision tree. What they look for first, what they let pass, what they always kick back, what they have learned to recognize over a career. We translate that into the rules the agents follow. Every subsequent file runs through those rules, with exceptions surfaced for human review and the judgment calls documented at the point of decision.

The practical effect is that the senior operator’s expertise stops being a single point of failure.

One of our community bank customers has been piloting Saris with a new hire as the first user. The concept is simple: when the agent carries the institution’s judgment, the new hire’s learning curve shrinks from months of absorbed pattern recognition to days of workflow fluency. The underwriting floor rises for every team member on day one, and keeps rising as the agent learns.

The other practical impact, less often discussed but almost equally valuable, is the one your risk and compliance teams will be overjoyed about. The reasoning behind every decision is now visible, logged, and auditable. That means come audit time, your team is ready. The reasoning is already captured rather than being reconstructed from memory after the fact.

The question from here

At this point, most community finance leaders understand that they need AI somewhere in their workflows. The open debate is now where AI can be best applied. Across customers, Saris is seeing significant impact in the specific operational knowledge your institution depends on. That knowledge belongs in a system of record that will outlast long past the current team.

outlThis is also a use case that depends on speed. Every quarter you wait is a quarter you compound exposure to the problem. The senior operator carrying the team’s judgment today will retire or take a new role eventually. The onboarding runway for their replacement will be measured in months and years, not weeks. The members and dealers will feel it in cycle time and accuracy.

The institutions that get this right stop looking for a replacement and start looking at this as an operational and systems problem.