Proximitty Is Building Autonomous Loan Servicing Teams for Banks That Cannot Hire Fast Enough

The Macro: Loan Servicing Is a People Problem Without Enough People

Business loan servicing is one of those back-office functions that is simultaneously critical and chronically understaffed. Banks, credit unions, and fintech lenders all need servicing teams to manage active loans: collecting payments, handling delinquencies, managing modifications, processing early payoffs, and dealing with the thousand small tasks that keep a loan portfolio healthy.

The problem is that this work is labor-intensive, requires domain expertise, and does not scale well. A servicing specialist can manage a finite number of loans. When the portfolio grows, you hire more specialists. When the labor market is tight, you cannot hire fast enough. When collection rates drop, you throw more people at the problem. It is a linear cost model applied to a business that needs to scale exponentially.

The loan servicing technology market has seen investment, but most products focus on the system of record: the software where loan data lives. LoanPro, Peach Finance, and Canopy handle the data layer. What they do not do is perform the actual servicing work: the outreach, the negotiation, the judgment calls about when to push for payment and when to offer flexibility.

That is the gap Proximitty is going after. Not the data layer. The execution layer. AI agents that do the servicing work itself.

The Micro: No-Code Browser Agents That Learn and Adapt

Proximitty builds unified lending data layers and personalized risk scoring models to power AI agents that handle loan servicing workflows. The agents learn and adapt dynamically, improving collection rates over time. The platform uses no-code browser agents, which means deployment and management do not require engineering resources from the bank.

The personalized risk scoring is interesting because it means the agents are not applying the same collection strategy to every borrower. They assess individual risk profiles and adjust their approach accordingly. A borrower who is two days late for the first time gets a different interaction than a borrower with a pattern of late payments. This kind of nuanced, borrower-specific servicing is exactly what human specialists do well and what traditional automation does badly.

The company claims to be working with four large institutions within three weeks of launch, which is remarkable traction for a fintech product selling into banking. Banks are famously slow to adopt new technology, and getting four signed within three weeks suggests either very strong founder relationships or a pain point so acute that normal procurement timelines go out the window.

Wye Yew Ho (CEO) and Zi Zhang (CTO) are the cofounders. Ho previously led FinCrime and Growth at Taptap Send and advised banks and fintechs on risk strategy at McKinsey. Zhang was a security infrastructure lead at Bloomberg and Head of Engineering at ACI.dev. The McKinsey and Bloomberg backgrounds are directly relevant: Ho understands how banks think about risk, and Zhang knows how to build enterprise-grade infrastructure. The company went through Y Combinator’s W26 batch.

The regulatory dimension is significant. Loan servicing is heavily regulated, and AI agents contacting borrowers need to comply with FDCPA, TCPA, and state-specific collection laws. Getting this wrong is not a product issue. It is a legal liability issue for the bank. Proximitty needs to have the compliance layer airtight from day one.

The Verdict

Proximitty is attacking a real bottleneck in financial services. Loan servicing does not scale with people, the people are hard to find, and the work is repetitive enough to be a strong fit for AI agents. The personalized risk scoring adds the kind of intelligence that makes the agents better than the average human servicing specialist, not just cheaper.

At 30 days: compliance. Are the AI agent interactions fully auditable and FDCPA-compliant? One regulatory violation from an AI agent would be catastrophic for a bank customer.

At 60 days: collection rate improvements. The claim that agents “dynamically learn and increase collection rates” needs to be backed by data. How much improvement? Over what baseline? With what borrower segments?

At 90 days: can the platform handle complex modifications and workouts? Routine collections are the easy part. Loan modifications, forbearance agreements, and workout plans require nuanced judgment. If the agents can handle those, the value proposition multiplies.

The McKinsey plus Bloomberg founding team is strong for this market. Banks trust people who have worked at places they already trust. I think Proximitty has the right positioning and the right team for a market that desperately needs automation.