withfinbox.ai is where FinBox is rethinking Indian loan origination for the AI era, with lenders who want direct input into what it becomes. Here’s where we’ve landed.
First touch on Day 1. Disbursal on Day 12. Somewhere in between, the customer went cold.
Your RMs spend days chasing documents. Credit teams waste bandwidth reviewing incomplete files. The dashboard says 70% automation, but the team is still manually chasing.
A process built for paper, moved onto screens, and never redesigned for what it became.
None of this is a people or technology problem.
It is a design problem.
Almost every vendor selling ‘AI for lending’ today is answering the wrong question: where can we plug an LLM into the process we already have? That produces chatbots, smarter OCR, and a mild efficiency story. It does not produce a different loan.
The question worth asking is different. If we designed origination today, knowing what AI can be trusted to do and where it fails, what would we remove from the flow entirely?
Long, complex review steps exist only because no system could be trusted to answer on its own. Similarly, entire escalation loops exist because the assistant was never taught how the underwriter thinks.
A new origination stack, powered by agentic AI.
AI agents handling judgment where judgment can be automated.
Hard rules where compliance demands them.
Tested against every edge case, before any of it touches a real borrower.
A very well-deserved milestone for the FinBox team. Our collaboration on embedded finance and digital lending has shown us the thoughtfulness and depth they bring to solving hard problems. With Atlas, they’re taking that further — putting agentic AI at the heart of loan origination. It’s an exciting development for the industry, and we look forward to seeing the impact it creates.
If that’s interesting, come and build with us. See what’s in progress, shape what goes in, and have a direct line to the people making the calls.
Start a conversation
10+ years of NBFC and HFC conversations. Knows where the process breaks because he’s seen it break the same way, at scale, across 200+ institutions. Writes about it at withfinbox.ai.
Building the AI infrastructure stack. Designed the synthetic testing framework. Obsessed with what reliability in regulated financial AI actually requires — and how far most vendors fall short.
We’re publishing field notes from inside the work — what we’re trusting AI to do, where it’s breaking, the calls we’re making as we design an agentic LOS in production. Real-time, unfiltered, straight from the people writing the code.