Fintech · AI Strategy · 2023

Financial Inclusion Platform

Designed trust-critical experiences for Rapipay and NYE Financial — banking for India's unbanked 100M users processing $3B+ in annual transaction value.

Scroll to explore

← Drag or scroll →
01

Problem

Trust Before Features

First-time users who have never trusted a bank need clarity, not complexity.

$3B+Annual transaction value
02

System

AI Trust Model

Built explainable AI patterns for credit, KYC, and support — transparency as product feature.

100MUnbanked users targeted
03

Process

Inclusive UX Research

Tested with tier-2/3 users, low-literacy flows, and agent-assisted onboarding.

20Weeks engagement
04

Outcome

Inclusion at Scale

Platform designed for first-time financial product adoption.

3Product lines unified
Client
Rapipay · NYE
Engagement
AI + Product Lead
Duration
20 Weeks
Primary Outcome
Inclusion at Scale
Stack / Tags
AI · Fintech · Trust UX

Artifacts from this engagement

ProblemChallenge BriefConstraints and context before the engagement ProcessEngagement PhasesHow the work was structured and delivered SystemSolution ArchitectureWhat was built and how it works OutcomesResults & MetricsMeasurable impact and learnings
Fintech / Inclusion / AI Strategy

Financial
Inclusion
Platform.

Designed the trust-critical product experience for Rapipay and NYE Financial — a fintech ecosystem bringing banking services to India's unbanked 100M users. $3B+ in annual gross transaction value processed through an interface built not for power users, but for first-time users who have never trusted a bank.

$0
Annual Gross Transaction Value
100M+
Target Users Reached
58%
Transaction Completion Rate ↑
2022
Year Completed
Client
Rapipay · NYE Financial
Role
AI Strategy + Product Design Lead
Timeline
20 Weeks
Outcome
Financial Inclusion at Scale
03
Section 01
The
Problem.
Designing for users who have never trusted a financial institution — and had good reasons not to.
A Product Built
for the Wrong User.

India's fintech ecosystem had expanded rapidly — but the 100M users at the bottom of the economic pyramid remained unreached. Not because the products didn't exist. Because they were designed by urban product teams for urban, digitally literate users. The unbanked user — often operating in a second or third language, often using a smartphone for the first time, often carrying deep institutional distrust of financial systems — was an afterthought.

Rapipay's ambition was different: to build financial infrastructure for Bharat, not India. The design brief was the hardest kind — not aesthetic refinement, but fundamental rethinking of what trust looks like in a digital product when the user's default assumption is that the system will work against them.

01
Transaction abandonment at 72% — users would initiate money transfers and stop before completion. Not UX confusion — active fear. The interface communicated risk, not safety, at every critical moment.
02
Language and literacy gap — the product was English-first in a market where over 60% of target users were more comfortable in Hindi, Bengali, or Tamil. Financial terminology was impenetrable to first-time users.
03
No AI-assisted error recovery — failed transactions produced cryptic error states with no guidance. For users with low digital confidence, a single error was a reason to abandon the product permanently.
04
Agent network not supported by the product — most users transacted through human agents, not directly. The product had no interface designed for the agent as an intermediary — a fundamental mismatch between the actual use model and the design model.
04
Section 02
The
Process.
20 weeks. Fieldwork-first. Nothing designed before the user was understood.

Fieldwork Before
Figma.

Five weeks of field research in Uttar Pradesh, Bihar, and Tamil Nadu before a single wireframe. Understanding how trust actually works for this user — not how designers imagined it — was the only foundation that could support what came after.

Key Insight · Week 4 · Lucknow Field Visit
Trust for this user is not built by the interface. It is built by the person — the local agent. The product's job is not to replace human trust. It is to extend it.
01
Field Research
Weeks 1–5

Embedded with agent networks in UP, Bihar, and Tamil Nadu. Observed 200+ real transactions. Mapped the trust architecture: who users trust, when, and why. Documented the exact moments where digital products lose users — and the agent behaviours that recover them.

02
AI Strategy Design
Weeks 6–9

Architected the AI layer: intent recognition for low-literacy input, multilingual error recovery, predictive transaction assistance, and fraud pattern detection that didn't penalise first-time users. The AI had to feel like a knowledgeable friend, not a compliance system.

03
Trust-Layer Design
Weeks 10–14

Designed the three trust layers: immediate (visual confirmation at every transaction step), agent-mediated (interface support for agent-as-intermediary model), and systemic (status transparency, reversibility signals, institutional credibility markers). Every screen answered one question: "Is it safe to continue?"

04
Multilingual System
Weeks 15–17

Built the multilingual content system for Hindi, Bengali, Tamil, and English — not translations but full language variants. Financial terminology rewritten for each language with non-literate users in mind. Voice-over integration scoped for future state.

05
Agent Interface + Launch
Weeks 18–20

Designed the dedicated agent interface — a second skin over the consumer product built for the agent use case. Faster transaction flows, customer account overview, error escalation, cash management. Tested with 40 agents across 3 states before launch.

05
Section 03
The
Solution.
Four design decisions that rebuilt trust from the first tap to the final confirmation.
01
01
Progressive Trust Architecture

Every transaction screen designed with a single trust question at its centre: "Is it safe to continue?" Visual confirmation at each step, reversibility signals before commitment, and explicit success confirmation designed for users who had never successfully completed a digital transaction before.

02
02
AI-Assisted Error Recovery

Failed transactions no longer produced cryptic error codes. The AI layer diagnosed the error type, explained it in plain language in the user's preferred language, and provided a clear resolution path. Error recovery rates increased from 14% to 61% in the first month post-launch.

03
03
Agent-Native Interface

A dedicated interface layer for the 50,000+ agent network — not an afterthought but a primary design surface. Optimised for agent workflows, customer account management, and the specific trust dynamics of human-intermediated digital transactions.

04
04
Multilingual Content System

Full language variants for Hindi, Bengali, Tamil, and English — not translations, but language-native rewrites of every financial concept. Financial terminology simplified to primary school reading level without condescension. The system spoke to users in the language they trusted.

Rapipay · Trust Layer Interface · Consumer + Agent Views
₹500
Send Amount
→
Recipient
Trust Signal — Step 2 of 3
AI Recovery Suggestion
47
Today's Txns
₹2.1L
Volume
Fig. 01 — Consumer transaction flow (trust architecture), AI error recovery layer, and agent dashboard. Three surfaces. One coherent trust model.
06
Section 04
The
Outcome.
What changed — in transaction volume, in trust, and in reach.
3B
$3
B+
Annual GTV
Gross transaction value processed through the platform
58
0%
Completion Rate ↑
Transaction completion rate improvement post-redesign
61
0%
Error Recovery Rate
Up from 14% pre-AI intervention. Failed txns now resolved, not abandoned.
50K
50K+
Agent Network
Agents operating the platform across 18 states
✦
Transaction abandonment dropped from 72% to 28% — the trust architecture resolved the fundamental fear state that was driving abandonment. Users who once stopped at confirmation now complete. The design change had direct revenue impact.
✦
Agent productivity increased 40% in the first quarter — the agent-native interface reduced transaction time per customer from 6 minutes to under 2 minutes. Agents could serve 3x more customers per day with the same effort.
✦
Multilingual adoption exceeded projections by 3x — within 90 days of launch, 67% of active users had switched to Hindi or Tamil as their primary language. The assumption that users would default to English was a product team bias, not user reality.
✦
First-time user retention at 6 months reached 71% — the target population for financial inclusion products historically has very low retention because early experience failures are catastrophic. The trust architecture held first-time users through their most vulnerable transactions.
"

Most fintech products are designed for users who already trust the system. Raghvendra designed for users who had every reason not to. The work he delivered was not a redesign — it was a fundamental rethinking of what trust means in a digital financial product. The GTV numbers are the proof, but the real outcome is the 100 million users who now have access to financial infrastructure that treats them with dignity.

Chief Product Officer
Rapipay · NYE Financial · 2022
07
Section 05
Key
Learnings.
What designing for the unbanked permanently changed about the approach to trust-critical interfaces.
01
01
Trust is the Product.

In financial inclusion, the functional product — money movement — is table stakes. The actual product is the emotional experience of safety. Every design decision must answer one question before any other: does this make the user feel that continuing is safe? When that question is not the primary frame, you build for yourself, not for them.

02
02
Field Research is Not Optional.

No amount of user persona work, secondary research, or empathy mapping replaces time in the field with real users in real contexts. The agent network insight — which shaped the entire architecture — was not discoverable through digital research methods. It required physical presence in Lucknow, in a store, watching a grandmother attempt her first transfer.

03
03
AI Must Humanise, Not Automate.

The AI integration in this product was not about efficiency — it was about empathy at scale. The AI recovery system succeeded because it was designed to sound like a knowledgeable, patient person — not a system response. AI that replicates a system's voice magnifies the system's distance. AI that replicates a human voice builds the bridge.

04
04
Language is not Translation.

Building a Hindi interface is not translating an English interface. The linguistic structure, the financial concepts, the tone, the implied relationship between product and user — all of these are different. The multilingual system succeeded because it was written from scratch in each language, not translated. That distinction made the 3x adoption premium possible.

← Previous Case Study
Scale
Engine
Verizon · Design Systems · 50M+ Users
←
Next Case Study →
Maritime
OS
Crowley Maritime · Logistics · Legacy Modernized
↗

Trust-Critical
Product?

AI Blueprint is the engagement for founders who need AI strategy that works for real users — not power users. Financial services, healthcare, inclusion products — 8 weeks, transparent pricing from ₹5L.

Initiate Inquiry ↗

Explore More
Work.

Eight more case studies across enterprise systems, design systems, design leadership, and product — all with the same structural rigour applied here.

View Selected Work ↗