Siddharth Gore

Siddharth Gore

UX/MOBILE APP · AI

UX/MOBILE APP · AI

LendClear

LendClear

Project brief

A loan journey that turns rejection into direction

A loan journey that turns rejection into direction

Why it matters

Why it matters

An honest answer before the effort. An alternative instead of a dead end. A way back after a no.

An honest answer before the effort. An alternative instead of a dead end. A way back after a no.

What changed

What changed

Analyzed 4,269 real loan applications. Approval hinged on one number — so three AI-assisted flows followed: an eligibility check, a counteroffer engine, and a recovery plan.

Analyzed 4,269 real loan applications. Approval hinged on one number — so three AI-assisted flows followed: an eligibility check, a counteroffer engine, and a recovery plan.

Timeline

1-2 Weeks

Process

Data analysis, secondary research, user flows, wireframing, critique

Tools

Claude

Excel

Figma

AI Guidelines

Overview

One credit-score threshold quietly decided who got a loan. Finding it turned a binary rejection into decision support.

One credit-score threshold quietly decided who got a loan. Finding it turned a binary rejection into decision support.

4,269

loan applications analyzed

550

CIBIL cliff discovered

25 min

saved from wasted effort

Evidence

The threshold that made a fair-looking form act binary.

The threshold that made a fair-looking form act binary.

The threshold that made a fair-looking form act binary.

01 — Approval is a cliff, not a slope.

01 — Approval is a cliff, not a slope.

Below CIBIL 550, 10.4% of applicants are approved. Above it, 99.7%. There is no slope between those bands.

Below CIBIL 550, 10.4% of applicants are approved. Above it, 99.7%. There is no slope between those bands.

02 — Term length was the only escape hatch.

02 — Term length was the only escape hatch.

For sub-550 applicants, 2-year terms cleared 47.6% and 4-year terms 56.8%. Above four years: zero approvals in 1,600 requests.


The same person, the same lender: refused at 12 years, a coin-flip at 4. The yes already existed.

For sub-550 applicants, 2-year terms cleared 47.6% and 4-year terms 56.8%. Above four years: zero approvals in 1,600 requests.


The same person, the same lender: refused at 12 years, a coin-flip at 4. The yes already existed.

03 — The form asks twelve questions the decision ignores. Graduates 62.5%, non-graduates 62.0%. Self-employed and salaried: 62.2% each.


1,785 people sat below the line. 1,600 completed the whole form anyway — 25 minutes for a decision already made.

03 — The form asks twelve questions the decision ignores. Graduates 62.5%, non-graduates 62.0%. Self-employed and salaried: 62.2% each.


1,785 people sat below the line. 1,600 completed the whole form anyway — 25 minutes for a decision already made.

Why it compounds: refused without a reason, people apply elsewhere. Each attempt is a new hard inquiry. CIBIL reports average drops near 50 points for multiple applications within six months.

Why it compounds: refused without a reason, people apply elsewhere. Each attempt is a new hard inquiry. CIBIL reports average drops near 50 points for multiple applications within six months.

Priya
25 minutes, 13 fields, rejected with no reason. She applies at three more lenders — each one costs her score.

Problem statement 1: How might we intervene before the effort, and turn rejection into recovery?

Rajesh
CIBIL 492, asks for 20 years. Flat no — though a 4-year term from his profile clears 57% of the time.

Problem statement 2: How might we surface the alternative the lender already approves?

Design principles

AI that explains itself without taking over

I wrote the AI a job description before drawing a single screen.

The AI does exactly three jobs:

  • Predict — estimate approval likelihood from a soft credit check, before any effort

  • Search — scan the lender’s own approval history for the nearest viable alternative offer

  • Personalize — turn a rejection into an individual recovery plan, not generic credit tips


And it follows a behavioral contract:

  • It appears at exactly two moments (before the decision, after the decision). No chatbot. Otherwise invisible.

  • Says “unlikely,” never “you will be rejected.” Even the lowest band clears 1 in 10.

  • Every verdict names its main reason and the distance to the goal.

  • If bureau data is unavailable, the normal application continues. A layer, never a gate.

I didn’t invent these rules. They operationalize Microsoft’s HAX guidelines, Google’s People + AI Guidebook, and the Shape of AI pattern library.

My position in one line: the best AI UX in lending is the kind you don’t see.

I didn’t invent these rules. They operationalize Microsoft’s HAX guidelines, Google’s People + AI Guidebook, and the Shape of AI pattern library.

My position in one line: the best AI UX in lending is the kind you don’t see.

Interface system

Three flows, three honest answers

Three flows, three honest answers

Flow 1 — The quick check. Three fields, one soft check, no mark on the score. Thirty seconds to a verdict in three bands — Strong, Possible, Unlikely — with the reason named. “Proceed anyway” is always available.

Flow 1 — The quick check. Three fields, one soft check, no mark on the score. Thirty seconds to a verdict in three bands — Strong, Possible, Unlikely — with the reason named. “Proceed anyway” is always available.

🟦 HAX G1–G2 · PAIR Mental Models · Shape of AI Transparency

🟦 HAX G1–G2 · PAIR Mental Models · Shape of AI Transparency

Flow 2 — The counteroffer. When the answer would be no, the engine finds the nearest yes in the lender’s own history. Both options arrive together: ₹26,660 a month over 20 years, unavailable, beside ₹64,300 over 4, likely approved. The EMI doubles; lifetime interest drops ₹33 lakh.

Flow 2 — The counteroffer. When the answer would be no, the engine finds the nearest yes in the lender’s own history. Both options arrive together: ₹26,660 a month over 20 years, unavailable, beside ₹64,300 over 4, likely approved. The EMI doubles; lifetime interest drops ₹33 lakh.

🟦 HAX G9 & G11 · PAIR Explainability + Trust

🟦 HAX G9 & G11 · PAIR Explainability + Trust

Flow 3 — The recovery plan. For a true not-today: 58 points to go, two ordered steps, and one warning against applying elsewhere right now. The primary button isn’t an exit — it’s “Remind me in 3 months.” A scheduled notification, never a background check.

Flow 3 — The recovery plan. For a true not-today: 58 points to go, two ordered steps, and one warning against applying elsewhere right now. The primary button isn’t an exit — it’s “Remind me in 3 months.” A scheduled notification, never a background check.

🟦 HAX G10–G11 · PAIR Errors + Graceful Failure

🟦 HAX G10–G11 · PAIR Errors + Graceful Failure

What the critique caught

What the critique caught

No usability tests in a one-week sprint. Every screen went through structured critique rounds instead.

  • A reassurance that outlived its truth. “This won’t affect your CIBIL score” sat under the button that starts a formal check. Replaced with “Continuing includes a formal credit check.”

  • A verdict that looked like a toggle. The “Possible” pill looked draggable. It became a three-band scale — you are in a band, not at a point.

  • My own numbers contradicting each other. A nudge said reduce tenure to 10 years while the counteroffer said 4.

  • A cross-sell in a vulnerable moment. Premium card offers on a screen for someone just refused a home loan. Deleted.

  1. How we would know it works

  1. How we would know it works

Wasted effort. 37% of applications were completed by people with effectively no chance. A 25-minute form becomes a 30-second answer.

Rejections turned into offers. Today every no is terminal. Half of short-tenure requests from this segment already get approved, so even 15–20% acceptance is net-new lending.

The trap, defused. Fewer hard inquiries within 30 days — and how many refused users return at the 3-month reminder.

One lender, one product, behavioural data without interviews. A jump from 0 to 100 between adjacent score bands is too clean for real underwriting. I treated it as a design probe: one lender’s revealed policy, not universal truth.

Thank you for viewing

Home

View Project

©2024 All Rights Reserved by Siddharth Gore Design

Create a free website with Framer, the website builder loved by startups, designers and agencies.