- Intake: The application, its documents and the applicant's identity, collected through the form you design.
- Underwriting: Specialist agents read the evidence and your credit policy, and reason to a recommendation.
- Decision: The agents recommend. A person with the right authority signs.
- Offer: The offer and closing steps your workflow sets out for the product.
- Funding: Security registered and confirmed by the registry before the loan is funded.
Galileo is an agentic loan origination system for banks, NBFCs and lenders.
Specialist AI agents read each application, its documents and your credit policy, and recommend an outcome with the evidence behind it. A person with the right authority signs.
Three steps on every application.
Reads
Galileo reads the application and its documents: identity, payslips, bank statements, accounts, invoices and valuations. It also reads the passages of your credit policy that are closest to the application.
Reasons
Specialist AI agents weigh the evidence, your rules and your scorecards, and reason to a recommendation. Every claim must cite evidence. A claim with no evidence is rejected.
Refers
The recommendation goes to a person with the right authority, who signs. Guardrails send a file to a person. That person can set the AI aside and decide the file themselves.
You design your loan products.
Set the terms, eligibility, pricing, fees, documents, application form, workflow, rules and scorecards yourself. Or describe the product in plain words and the design assistant drafts it.
- The assistant asks only what it cannot work out from your words.
- Every change is a new version.
- A second person approves each version before it serves applications.
Your credit policy, read on every application.
Your credit policy is loaded into a policy library and split into passages. Each underwriting run is given the passages closest to the application.
- Search the library with a question and see the passage that applies.
- Each rule states its reason in plain words.
- Each rule cites the policy section it follows.
Agents reason from the evidence in the file.
Specialist AI agents read the application, its documents, your policy and your rules, and reason to a recommendation. The reasons name the policy rules they rest on.
- Every claim must cite evidence.
- A claim with no evidence is rejected.
- Rules and scorecards are inputs the agents weigh.
The agents recommend. A person signs.
Guardrails send a file to a person. A person with the right authority signs the decision, within approval limits you set by product and amount.
- The decision card shows what the agents recommend and why.
- The person signing can set the AI aside and decide the file themselves.
- Each signature is recorded on the decision.
Every decision can be replayed step by step.
Each decision is recorded with its evidence, reasons, policy version and signatures. The agents' recorded reasoning can be played back step by step.
- A full audit trail for every application.
- The policy version in force is recorded with the decision.
- The record can be downloaded.
