The AI-native LOS.
Money in the bank in about ten minutes.
Sentinel prices every borrower, vision AI checks every selfie, and an ordinary loan runs from mobile number to payout with no human in the loop. Your team handles only the exceptions. RBI Digital Lending ready. Free to use.
The offer comes before the paperwork
Sentinel prices the borrower before a single document is fetched. Paid checks only for borrowers who said yes.
- A few short screens, then your rules and Sentinel's AI make the credit decision and price the offer
- Amount and tenure on a slider, with the exact repayment, fee and GST before accepting
- DigiLocker, selfie and bank-statement checks run only after the borrower accepts
- Sentinel re-prices the offer on real income and banking behaviour once statements arrive
Thirteen steps. About ten minutes. AI where it counts.
The mobile-first journey a borrower sees on a LendPro-powered lender's site. Steps where AI makes the call are marked. Tap a screen. Every screen is illustrative.
Steps 4–6 · Credit decision and offer. AI · Sentinel prices the offer Bureau consent on its own screen, then the bureau pull and your rules engine in one step. Sentinel prices the borrowers who pass. The borrower picks amount and tenure — 7, 15, 30 or 45 days — on a slider, and sees the exact repayment, fee and GST before accepting.
- Resume on any device
- Safe hand-offs to DigiLocker, AA, eSign and mandate pages
- Plain declined, paused and under-review screens
- Stuck-screen watchdog
- One open loan per borrower
- reCAPTCHA and OTP rate limits
The ordinary loan runs itself. Your team works the exceptions.
Your policy, not ours: your rules decide who qualifies, and Sentinel prices the borrowers who pass.
- Ordinary loans go from sign-in to payout with no human in the loop
- Referrals — AML hits, face-match doubts, employment questions — go to a review queue with maker-checker approval
- A versioned rules engine you control: approve, reject or refer, and change policy without a code release
- Every decision records the policy version that made it, in a complete, timestamped audit trail
- Signed inBorrowerWhatsApp OTP · Borrower ID and Application ID issued
- ApprovedRules enginePolicy v12 · bureau consent on recordWhy: Age, income and bureau history within policy
- Offer pricedSentinel₹25,000 · 30-day bullet · risk band recorded
- KFS acknowledgedBorrowerConsent text hash stored · payout account locked
- Agreement signedBorrowerAadhaar eSign · sealed copy stored
- Payout releasedSystemeNACH active · verified account · handed to LMS
An ordinary loan, sign-in to payout, with no human in the loop — and every step on the record.
Sentinel: the AI underwriter inside
Not a rules engine alone. Sentinel is LendPro's pricing layer — a model trained on 33M loan accounts and 1.5B repayment events. Nothing to integrate, nothing to configure on day one.
Not one flat ticket. Sentinel estimates the default probability at every amount and returns an affordable amount, a rate and a risk band for each borrower.
Catch the defaulters your rules pass, approve the safe borrowers a cutoff turns away, and price your best borrowers down.
Reads the month-by-month repayment sequence behind the score, and cash flow — not one bureau score and a checklist.
Every decision comes with the reasons behind it, from a monotonic, explainable model your credit team can question.
Every rate traces back to a default probability — the risk-pricing derivation the Fair Practices Code asks lenders to justify.
Each model version is tested before release and ships behind the same endpoint. Nothing changes on your side.
Sentinel keeps scoring after disbursal and flags risk months earlier — while you can still act on it, not in the NPA report.
In your first month your repayments join the loop. Your curve is fitted to your book, then kept current.
Every decision produces an outcome and every outcome becomes training data. A rules engine is as smart on day 300 as on day one.
Identity verified by source — and by vision AI
Every fact is checked on the server against a verified source, never trusted from the borrower's phone. Vision AI confirms a live person who matches the Aadhaar photo, and smart name matching handles initials and order. A check that could not run is recorded as “not run” and never declines a genuine borrower.
| Check | What it proves | If it fails |
|---|---|---|
| PAN registry lookup | The PAN exists in income-tax records; name and date of birth come from the registry, not typed | Borrower corrects the PAN, with limited attempts |
| PAN lock | Once confirmed, a PAN can never be swapped on that application | Swap refused |
| AML / PEP / sanctions | The borrower is not on sanctions or politically-exposed lists | Sent to manual review |
| Employment (EPFO) | Employer and employment match what was declared | Decline or manual review, per your policy |
| DigiLocker Aadhaar | Government-issued identity: name, date of birth, address, photo | The journey cannot continue without it |
| Aadhaar vs declared details | PAN, date of birth and name agree — smart Indian-name matching handles initials and order | Mismatch declines or refers |
| Liveness selfie | A real person is in front of the camera, not a photo | Retry |
| Face match | The selfie matches the Aadhaar photograph | Manual review — never an automatic decline |
| Bank penny drop | The account exists and the holder's name matches | Re-enter the account |
| Re-use checks | The same PAN, Aadhaar or bank account is not reused across borrowers | Refused or referred |
| CKYC lookup | Central KYC record checked at disbursal | Recorded |
- Shadow mode to tune, strict mode to act
- Aadhaar held masked
- Attempt limits on PAN, DigiLocker and penny drop
- Manual review desk with maker-checker
Self-healing journeys
If a vendor is down, the journey pauses safely and resumes exactly where it left off. A watchdog re-checks any stuck step and moves the borrower on.
- Progress is saved on the server — switch phones or come back days later, and land on the same step
- The borrower is told when to come back; if the bureau or pricing is down, their daily try is refunded
- Every amount is calculated by one engine, in paise, so the offer, the KFS and the payout always match
- Every step is idempotent — a double tap or a retry can never pay twice
Humans in the loop, where it counts
One console for credit, operations and compliance: the exceptions the system routes to your people, and full visibility of everything it does on its own. Named-operator sign-in, masked mobile numbers, every read of borrower data logged, and live settings that change without a release.
Let AI agents do the analysis
LendPro is coming to MCP. Connect Claude or any MCP-compatible agent to your lending book, and hand it the work.
- Pull the funnel by source, offers by risk band, or vendor spend against budget
- Ask why approvals dropped yesterday, and get the rule and policy version behind it
- Have it draft a policy change — a person still approves it in the console
- Built on the same controls as the console: borrower data masked, every read logged, maker-checker on anything that moves money or policy
Why did approvals drop yesterday?
Approvals fell from 58% to 44% after policy v13 went live at 14:05. Most of the extra rejections came from the new income floor for self-employed borrowers.
Draft a fix for review.
Drafted policy v14, restoring the previous floor for self-employed borrowers. It is waiting for maker-checker approval in the console.
The agent proposes. A named person approves.
AI you can explain to the regulator
Built around the RBI Digital Lending Directions and the DPDP Act, with the reasons behind every Sentinel decision on record. Hosted in India.
The full cost on one page, generated by the same engine as the offer, acknowledged before signing.
Each consent on its own screen, recorded with a server timestamp and a hash of the exact text.
Shown to the borrower inside the app.
The borrower sees the window and exactly what to return.
Versioned consent text, and when each borrower accepted which version.
Sensitive data expires on schedule.
Handled from the console.
Aadhaar via DigiLocker, PEP and sanctions screening, CKYC lookup.
The borrower app holds no authority — every decision and every amount comes from the server.
Selfies, statements, credit files, agreements and bank details.
Vendor keys never reach the browser; logs carry no borrower personal data.
Money goes only to the bank account the borrower verified.
Every rail an AI-native lender needs, already wired in
Each vendor sits behind a standard adapter, so adding or switching one never changes the journey. Every call has timeouts, safe retries, circuit breakers and a spend budget, and every vendor can run in mock mode for demos and testing.
An AI-native lending stack, live on your brand in a day
Bring the lending licence and the capital. LendPro brings the stack.
- White-label: brand, colours and journey screens set per lender
- Runs on AWS Mumbai behind a web application firewall — data stays in India
- Every disbursed loan is handed to LendPro LMS automatically
- Mock mode for every vendor, so you can demo and test before going live
LendPro AI is free
There is no licence fee.
- The full LOS, with LendPro LMS for every disbursed loan
- Sentinel's AI pricing on every offer
- The operations console for your credit, operations and compliance teams
- Pre-integrated KYC, bank, bureau, eSign and mandate partners
Put AI to work on your book.
Tell us about your licence and your product, or call the LendPro team directly.