01 / 16 Use ← → to navigate
Founder Office · Assignment Response

I am comparing Loans24 with a gold loan company.

By slide 10, you will agree it was the right call. The reason a used car loan and a gold loan look like siblings is the same reason Loans24 can ship the fastest secured lending stack in India.

Divyanshu Verma IIT Roorkee · Economics· Product, Fintech
SECURED
FIELD-OPS
SUB 30 MIN
LOANS24
RUPEEK
02 · The Mirror Thesis

Everyone benchmarks against a bank.
The smarter mirror is Rupeek.

A bank shows you what scale looks like. Rupeek shows you what speed looks like when the asset itself is the underwriting. Loans24 is the second kind, not the first.

Trait Loans24 Rupeek Match
Collateral is the moat Used car valuation Gold purity test
Borrower profile Self employed, thin file Self employed, thin file
Field operations Dealer touchpoint, RC check Doorstep agent, gold assay
Decision speed promised Loan offers in minutes 30 minute disbursal
Ticket size sweet spot Rs 3L to Rs 15L Rs 50K to Rs 10L
Repeat behaviour Top up against same car Re pledge same gold
Source · Rupeek company materials, Cars24 / Loans24 public pages, RBI co lending guidelines.
03 · The Uncomfortable Truth

Loans24 holds the better asset. And still moves slower.

Rupeek lends against gold it knows nothing about until an agent reaches the door. Loans24 lends against a car that Cars24 has already inspected, valued and tracked. Yet the heavier, better known asset gets the slower answer.

Heavier asset
Loans24 (Loan Against Car)
Car worth Rs 7.2L + full Cars24 data trail
up to 72 hrs
Lighter asset
Rupeek (Gold Loan)
Gold, no prior data, assayed on the spot
about 45 mins

This is not a product problem. It is a data infrastructure problem. And the data to fix it already lives inside the Cars24 ecosystem.

04 · The Borrower

Meet Mayank.

He runs a kirana store in Hadapsar, Pune. Two employees, one daughter in school, monthly UPI turnover of about Rs 2.1 lakh, but his actual take home after shop expenses is closer to Rs 65,000 once you count it the way a bank does. On paper he is invisible. ITR not filed for two years. He wants a 2019 Swift Dzire from Cars24 for Rs 5.2L.

Seven lenders have already said no in three days. He is the customer no one builds for. Loans24 should.

Self employed
70% of used car buyers
Thin file
No ITR last 2 yrs
Digital trail
UPI and GPay heavy
Mayank Pawar
Mayank Pawar, 34
Hadapsar, Pune
7 rejections
Monthly UPI inflow
Rs 2.1L
turnover, not take home, over 480 transactions
Declared income (ITR)
Nil
last filed FY22
Wants
2019 Swift Dzire VXi · LTV needed 80% on Rs 5.2L
What the system sees today
No CIBIL hitThin file flagManual reviewPending FOIRReapply later
05 · Side by Side Underwriting

One asks for paper. The other tests the asset.

Customer step System step Human / friction step
LLoans24 · today
~72 hr
App opened0 min
KYC + bureau pull8 min
Income proof upload (bank, ITR, GST)+30 min
Vehicle inspection slot booked+30 hr
Inspection report uploaded+15 hr
Underwriter queue+18 hr
Committee / exception+6 hr
Offer, e-sign, disbursal+2 hr
RRupeek · today
~45 min
App opened0 min
Doorstep agent dispatched (geo)2 min
Agent reaches home+25 min
Gold purity tested (XRF)+8 min
Instant value + LTV offer+2 min
e-KYC + e-sign+5 min
Disbursal to bank+3 min
Rupeek asks for no income proof. The gold is the proof.
Red rows are friction points where AI as core infrastructure changes the answer. The next slides expand on each.
06 · Where The Time Goes

The 72 hours hide where the time really dies.

KYC + bureau pull8 min
Income proof (bank, ITR, GST)30 min
Vehicle inspection slot booked30 hr
Inspection report uploaded15 hr
Underwriter queue18 hr
Committee / exception6 hr
Offer, e-sign, disbursal2 hr
Same seven steps as slide 5. Bars are proportional. Adds up to roughly 72 hours.
The document wall

Seven documents. Six month bank PDFs. Two year ITR. Around 70% of self employed drop off at this single step. Fix: UPI + GST + Account Aggregator consent reads income from source, in seconds.

Solvable in 0 to 3 months
The ghost inspector

Loans24 sends a field agent to inspect a car that Cars24 may already have inspected and stored. Fix: pull the condition score and live value from the Cars24 vehicle DB via API. No agent needed.

Solvable in 0 to 3 months
The sequential pipeline

Bureau, income, valuation and fraud checks run one after another. Each one waits for the last. Fix: fire all four checks the moment the form is submitted.

Solvable in 3 to 6 months
07 · Friction Heatmap

Seven places the journey bleeds time.

Each red dot is a step where a human is doing what software should do. Below each one is the AI substitution.

1
2
3
4
5
6
7
App openedUnderwritingDisbursalServicing
1Income proof chase
UPI + GST + AA inference
2Co applicant tug
Optional via cash flow score
3Inspection wait
Cars24 vehicle DB, no agent
4Underwriter queue
Parallel agent grid
5Fraud surfaces late
Pre cog at form load
6Disbursal handover
Single ledger event, no excel
7Post disbursal silence
Memory layer per borrower
08 · Cosmetic vs Infrastructure

Is your AI cosmetic or is it infrastructure?

Most lenders use AI to look modern. Very few use it to make the actual decision. The job description draws this exact line.

WATERLINE
Cosmetic AI · what most lenders ship today
Chatbot on the website Auto email after applying Credit score shown in app Colour coded dashboard
Infrastructure AI · what actually changes the outcome
Account Aggregator, replaces PDF uploadML underwriting, replaces first pass human reviewParallel pipeline, replaces serial queueVehicle intelligence API, replaces physical inspectionBorrower memory graph, replaces blank slate re verifyFraud signal layer, catches it before underwriting
The JD says it plainly: "The bar is not whether you use AI. It is whether the systems you ship decide and act." Everything in this deck sits below the waterline.
09 · The Unfired Gun

Loans24 sits on a gold mine. The pipes are not connected yet.

Every car that passed through Cars24 has a data footprint. Loans24 can reach all of it. But underwriting is not wired to it. So Loans24 sends an agent to inspect a car Cars24 already inspected in detail.

200 point physical inspection report
Live market value from auctions, 130+ cities
RC transfer records and ownership chain
Depreciation curve for the exact model and variant
Comparable sold cars, same city, last 30 days
Today: the field agent visits the car and asks questions Cars24 already answered when the car changed hands.
Should be: RC number entered. Cars24 API returns vehicle intelligence in 3 seconds. LTV calculated instantly. No agent. No wait. No cost.
200 pt Inspection Live Value Mileage +Condition RC + Ownership Depreciation(variant) Comparablesales 30 days Cars24 Vehicle DB DISCONNECTED today LOANS24 Underwriting System Should be: CONNECTED VIA INTERNAL API 3 second vehicle intelligence. No agent.
The punchline: Rupeek must send an agent to assay gold every single time. Loans24's asset was already assayed, the day Cars24 bought the car. We can be faster than Rupeek without anyone leaving the office.
10 · The Unlock

What I would build. In this order. Systems, not features.

The bar is not whether we use AI. It is whether the systems we ship can decide and act. These four do.

System 1 · Months 0 to 3 Doc wall killer

AA Powered Income Engine

Read income from source, not from PDFs

One Account Aggregator consent pulls verified bank cash flow and GST turnover. ML reads 12 month seasonality and EMI load. Replaces ITR, bank PDFs and GST cert upload. No human under Rs 5L. The same engine also runs fraud pre cog checks the moment consent is given, so obvious fraud never reaches an underwriter.

System 2 · Months 0 to 3 Inspection killer

Vehicle Intelligence API

RC in, value out, 3 seconds

Internal API connecting Loans24 underwriting to the Cars24 vehicle DB. Trigger on RC entry. Returns market value, condition score, history and comparables. Calculates LTV instantly and flags claimed vs real condition.

System 3 · Months 3 to 6 Queue killer

Parallel Underwriting Grid

Six workers, not one queue

Bureau, AA income, vehicle valuation, fraud, FOIR and co lender eligibility and same day co lending reconciliation run as parallel agents. Auto approve when all thresholds pass. Human reviews only the exceptions. 8 to 10 files a day becomes 50+.

System 4 · Months 6 to 9 Cold call killer

Memory Layer for Collections

Every conversation feeds the next

Voice and chat transcripts become a borrower keyed store. The next agent, human or AI, opens the call already knowing the last promise to pay. The same memory powers top up, cross sell and pre closure.

11 · System Wireframe

The architecture, end to end.

Event bus in the middle. Agents on the sides. A memory store under everything. No service waits on a human queue.

CustomerApp / WhatsApp / Dealer Ingest Layerconsent · OCR · AA EVENT BUSidempotent · replayable Pre cogSMS+Bureau+Vehicle Underwriting Grid6 parallel workers Fraud Pre cogdevice + KYC graph Co lend Recon3 way match Memory Writervector store Decision Enginepolicy + ML DisbursalNACH · partner pool Collections AImemory aware DealerOSlistings · payouts Memory Store · Postgres + Vector DB
12 · The 15 Minute Loan

Fifteen minutes is not a slogan. It is an architecture.

When the heavy work happens before the customer hits submit, and the workers run in parallel, the wall clock collapses. Here is how the minutes are spent.

Customer intake3 min
Pre cog parallel pull2 min
Underwriting grid4 min
e-KYC + e-sign3 min
Disbursal NACH3 min
Total wall clock15 min
END TO END
0
minutes
vs 72 hr today
13 · The Compounding Moat

The 90 second loan that only Loans24 can build.

Memory based underwriting for repeat borrowers. The moat that no bank and no pure play NBFC can copy.

Today, when Mayank returns for a second loan: the system asks for all seven documents again. A field agent schedules another inspection. The underwriter reviews from scratch. Same 72 hours.
With the memory layer: Mayank opens the app. System recognises him. Pulls his last underwriting packet. Runs delta checks only. CIBIL changed? No. GST turnover up 8%. LTV still safe? Yes. Car condition? Cars24 API confirms. Decision: approved. EMI shown.
Loan 1 behaviour GST trajectory Vehicle condition(Cars24 API) Bank cash flow (AA) Bureau history Businessseasonality MayankBorrower Node Underwriting Decision Engine Every loan makes the next one easier At month 12, Mayank's decision takes 90 seconds, not 90 minutes
Rupeek can build a borrower memory. Loans24 can build a borrower plus vehicle memory. No bank, no NBFC, no fintech in India has both in one system. That is the category moat.
14 · Projected Impact

What the numbers look like in two quarters.

Conservative estimates, modelled on Rupeek gold loan ops benchmarks, the Account Aggregator ecosystem and the Cars24 dealer footprint. Happy to walk through the working live.

Underwriting TAT
0min
288× faster · from 72 hours
Thin file approval
0%
+27 pts · from 18% today
Cost per disbursal
0idx
−38% run rate
Collections contact
0%
memory aware dialer
Monthly disbursal
Rs 250 Cr to Rs 420 Cr
Same headcount, same NIM
Self employed share
32% to 55%
Thin file becomes a feature
Co lending recon
T+3 to T+0
Partners see settled book daily
At Rs 250 Cr a month, even a 10% move in any of these metrics is real money. These are outcomes of three buildable systems, not aspirations.
15 · The Prototype

What the ideal flow looks like for Mayank.

Five screens. Under 15 minutes. No agent. No PDF. No waiting. Built on the four systems from slide 10.

9:41
Hi Mayank!
Tap to start your loan against car.
2019 Swift Dzire VXi
2019 Swift Dzire VXi
Est value Rs 5.2L
Cars24 Verified
Check Eligibility
Step 1 · 30 sec
9:41
Share your financial story
Takes 10 seconds.
Bank cash flow
GST + UPI inflow
One tap. No PDF. Account Aggregator.
Allow and Continue
Step 2 · 90 sec
9:41
Analysing your profile
Cash flow verified Rs 2.1L
Cash flow score: strong
Vehicle value Rs 5.2L
Fraud signals clear
No action needed.
Step 3 · 3 to 4 min
9:41
Congrats!
You are eligible for
Rs 4,16,000
At 13.5% p.a.
EMI Rs 9,560 / month
Tenure 60 months
Accept Offer
Step 4 · Instant
9:41
Loan being processed
Offer accepted
Agreement signed
NACH mandate set
Rs 4,16,000 in your account by 3:45 PM today
Step 5 · 8 to 12 min
Total: 12 to 15 minutes · first time borrower · zero physical touchpoints
16 · Why Me, Briefly

I have already built three of the four primitives Loans24 needs. Let me build the fourth with you.

Worthy Advisors
Treasury automation
60+ hours of manual ops removed every month.
Worthy Advisors
Bank reconciliation engine
98% match accuracy across 3 lakh entries.
Self shipped product
RUPAIYA, AI money companion
30% conversion at 3:1 LTV to CAC.
Portfolio preview
SEE MY FULL PORTFOLIO
itsdivyanshu.me
Live case studies, shipped products and writing built end to end. Click to open in a new tab.
Visit portfolio
CALL TO ACTION
Give me the underwriting problem. I will
hand back a working prototype in one week.
A Loom walkthrough of this deck plus a working demo of the Pre cog engine skeleton is ready to share.
divyanshu.iitroorkee@gmail.com +91 97724 27180
Loom link in submission thread
DIVYANSHU VERMA · IIT ROORKEE · FOR NEHAL SIROYA · LOANS24 BY CARS24
Use the arrow keys or click the dots to move
9:41100%
loans24
Screen 1 of 6