90 minutes to redesign
India's largest bank.

Chapter 1 · The full UX audit

Every design choice prioritises the institution
over the customer.

SBI's net banking requires navigating four pages across three subdomains just to log in. Each page has a different visual identity. The login page disables copy-paste on the password field. The CAPTCHA fails on correct input frequently enough that users have learned to expect it. The OTP page has a wall of security text that nobody reads.

The YONO dashboard presents every product SBI sells — deposits, loans, investments, cards, insurance — on a single page with no hierarchy. A scrolling carousel of cartoon icons sits above a relationship overview that shows account numbers but not balances. The right sidebar contains 24+ links. Promotional banners compete with functional elements. The interface is not outdated — it is actively hostile.

Chapter 2 · Why this is corruption

Procurement, accountability, indifference.

SBI's failure is not incompetence. India does not lack talented UX designers or world-class engineers. India built UPI — the largest real-time payment system on earth. The technology exists. The talent exists. The money exists ($1.3 billion in annual IT spending).

This is structural corruption. Technology contracts awarded through lowest-cost bidding, which systematically selects for the cheapest vendor, not the best design. Nobody at SBI is measured on user experience — the chairman reports net profit, the CTO reports uptime. Nobody reports how many customers gave up trying to log in today.

The people making decisions at SBI do not use SBI's net banking the way their 520 million customers do. That insulation — the comfort of never having to use what you build — is its own form of corruption.

Chapter 3 · The infrastructure-interface paradox

UPI processes 228 billion transactions a year.
PSU banks can't build a login screen.

The IMF calls UPI the world's most advanced real-time payment system. Google Pay, PhonePe, and Paytm built beautiful interfaces on top of UPI infrastructure. Indian consumers proved they can and will adopt digital banking — when the interface respects them.

Yet every public sector bank — SBI, PNB, Bank of Baroda, Canara Bank — delivers interfaces that look like they were built in 2008. BOB's ‘bob World’ app was so badly managed that staff committed fraud to inflate registration numbers. The RBI banned them from onboarding new users.

The same government that built the world's best payment infrastructure cannot build a login screen. The difference is not capability — it is accountability. UPI had NPCI. PSU banks have procurement committees.

Chapter 4 · Each decision in detail

Every choice is traceable
to a real user behaviour.

Amount chips — lifted from real transaction history. Last paid ₹34,000 on 30 Mar. The user almost never types.

Removing the CAPTCHA was the first decision. CAPTCHAs exist to prevent bot attacks, but SBI already requires OTP verification. The CAPTCHA adds friction for every human user to solve a problem that OTP already solves. It is security theatre — visible effort that provides no additional security.

The amount chips came from lived experience. When paying rent monthly, the amount is the same. The system has this data. Surfacing recent amounts as one-tap chips eliminates the most error-prone step in the payment flow — typing a large number on a phone keyboard.

Inline OTP — keeping verification on the same page as the action — is an engineering decision serving a design goal. In SBI's flow, OTP opens a modal or navigates to a new page. The user loses context. In the redesign, OTP slides in below the review summary. The user sees their payment details while entering the code. Only happens when designer and developer are the same person.

Chapter 5 · What AI-assisted development actually looks like

Claude executes.
The thinking is entirely human.

✔ Project initialised — 0:42
> build login flow with split-screen + animated OTP state
✔ components/login-form.tsx
✔ components/otp-input.tsx
✔ Framer Motion state transition wired
> build dashboard — balance, 3 actions, transactions, card
✔ app/dashboard/page.tsx + 7 child components
> build payment flow — chips from history, inline OTP
✔ 5-screen stepper, reusable across send-money + bills
$ git push origin main
✔ Vercel deploying...
✔ Live — sbi-redesign.vercel.app
# elapsed: 1h 28m

Inside the IDE

Real codebase. Real Next.js project. Hot-reload terminal running. AI accelerates implementation, not architecture.

This is not “describe an app and let AI generate it.” Every design decision — removing CAPTCHA, elevating three primary actions, using transaction-history chips — came from a decade of UX practice and lived experience as an SBI customer. The AI accelerated the execution. The thinking was entirely human. Designer and developer fused; AI accelerated both roles simultaneously.

Chapter 6 · What the research says

The data confirms what every SBI customer
already knows.

SBI Secure OTP app: 1.2 out of 5 stars. MouthShut rating for SBI net banking: 1.72 out of 5. IsItDownRightNow user rating: 1.6 out of 5. These are not outlier reviews — they represent the consistent experience of millions of users.

YONO claims a valuation of $40–50 billion, but this figure has never been independently validated. SBI added 164 ‘enhancements’ to YONO in FY2025 — none of which addressed the fundamental UX problems visible in the login flow. The institution knows. They just aren't accountable for fixing it.

Chapter 7 · What I'd build next

Three production-scope additions
for the next iteration.

The 90-minute prototype proved the design and the speed. A production-ready version would take weeks — because the hardest problem (understanding what users actually need) is already solved. The remaining work is engineering scope, not design uncertainty.

Chapter 8 · Mockup vs prototype

The difference between an opinion
and evidence.

Behance and Dribbble have dozens of SBI redesign concepts. Most are Figma mockups — beautiful screenshots that were never built, never tested, never deployed. They demonstrate visual design skill but not product thinking or engineering capability.

This redesign is working code. It runs in a browser. It is deployed on Vercel. You can click through the login, navigate the dashboard, and complete a payment flow. It demonstrates not just what a better SBI could look like, but that one person can actually build it — in 90 minutes.