PIX: payment flows
The PIX payment experience for one of Brazil's largest digital banks, used by millions of people. It covers everything from scanning a QR code to approving high-value transfers.
- Role
- UX Designer (Mid-level)
- Timeline
- Platform
- Mobile (iOS + Android)
- Team
- UX Designer · Product Manager · Engineering · Compliance · Business stakeholders
Overview
Banco Inter needed PIX, Brazil's instant payment system, across every use case: static and dynamic QR codes, copy-paste keys, high-value transfers with biometrics and an approval flow for business accounts making large payments. I was the UX designer responsible for mapping and designing these flows, in a team of 50+ designers, with tight deadlines and constant alignment with product, engineering, compliance and business stakeholders.
The problem
PIX launched in November 2020 and became mandatory for regulated institutions almost right away. For Banco Inter, with millions of active users, implementing PIX was a regulatory obligation with a fixed date. It was also a trust issue, since people had to feel safe moving money instantly, and a design challenge at a very large scale.
Each use case had its own flow logic:
- QR code scanning: static QR codes with a fixed amount, and dynamic QR codes with due dates, interest and late fees
- PIX copy-paste: people copy a PIX key in another app and paste it into Inter
- High-value transfers (individual accounts): limit checks, biometric authentication and date scheduling
- High-value transfers (business accounts with approval): large transfers need a second person to approve them, only during business hours
The hard part wasn't only technical. PIX is instant and can't be undone, so nobody tolerated an ambiguous confirmation or a poorly explained error.
How do we design for payments that can't be undone, at scale, in cases ranging from "scan and pay" to "get the CFO to approve this before the end of the day"?
Briefing
In a team of 50+ designers, briefing was a structured, layered process, very different from the lean discovery I was used to. I joined after product and compliance had already set the overall scope. I got a document with the functional requirements and the Central Bank's regulatory constraints, and took part in alignment sessions with product, engineering and legal before any design work began.
The main design requirements:
- QR flows: support every QR type defined by the Central Bank (static and dynamic). Dynamic QRs carry due date, interest and late fees, and all of it had to be clear on the review screen.
- QR scanning errors: camera permission denied, expired QR, invalid QR and unreadable QR. Each needed its own state instead of a generic error.
- PIX copy-paste: the pasted key had to be validated on the server and the result shown clearly, since people had no QR code to double-check.
- High-value, individual accounts: biometrics required above a threshold, with the available limit on screen and an option to schedule transfers that exceeded the daily limit.
- High-value, business accounts: approval required. The person starting the transfer makes a request, and a second authorized person at the company approves or rejects it during business hours. Requests made after hours had to be queued, not rejected.
The constraints were strict. Inter's design system had to be followed exactly, information density had to work across a wide range of devices, and the flows had to be production-ready on a short timeline because of the regulatory deadline.
Ideation & brainstorming
With scope and constraints clear, exploration was faster and more focused than on a greenfield product. The Central Bank's rules already defined much of the information hierarchy (the content of the dynamic QR review screen, for example). What they didn't define was how to present it.
The questions I explored:
- QR flows: how do you explain the difference between static and dynamic QRs to people who don't know or care about it? You don't. The review screen adapts and shows only the fields that QR type has. Due date and late fee appear only when the QR includes them.
- Bottom sheets: when to use a bottom sheet for confirmation and review steps without making the experience feel fragmented or full of modals.
- QR scanning errors: a denied camera and an expired QR are very different problems. People can fix the first on the spot; the second means contacting whoever they're paying. The errors needed visual hierarchy and clear instructions.
- Business account approval: how do you make an asynchronous, multi-person approval feel coherent on a phone? The person who starts it sends the request and waits. The approver gets a notification. Both needed a clear view of the status.
- Business hours: what happens to a business transfer request made after hours? Blocking it felt wrong. The answer was queuing it with a clear message: "your request will be sent to the approver on the next business day."
Lo-fi usability testing
With a short deadline and a big team, lo-fi testing at Inter was more structured than at smaller companies. Test plans were submitted and approved, sessions were coordinated with the research team, and findings went into a repository shared across teams.
I took part in the tests for the QR code flows and the business approval flow, the scenarios with the highest risk of user error and lost trust.
What the QR code tests showed:
- People who scanned an expired QR expected to see the problem in the camera view itself and didn't look at the bottom of the screen, where the error was. The warning had to appear inside the camera area, or the camera had to stop with a very clear interruption.
- The dynamic QR review caused hesitation: people weren't sure whether interest ("juros") and late fees ("multa") were being charged right now or were just information. Changing "juros" to "juros por atraso" (interest for late payment) cut the confusion a lot.
- With copy-paste, many people pasted the key and tapped "continue" without reading the review. The payee's name and bank needed visual weight, not a form row.
What the business approval tests showed:
- People who started a transfer only realized it was pending after leaving the screen and coming back. That's why a persistent "awaiting approval" card was added to the home screen.
- The business hours rule wasn't understood: people thought after-hours transfers had been rejected, not queued. The copy had to spell it out: "your request will be sent to the approver on the next business day."
- Approvers wanted more in the push notification (amount and payee name) to decide whether to open the app right away.
Flow refinement
With the results in, I revised the flows on both fronts before the final design pass.
QR code flow changes
Error states redesigned: expired and invalid QR errors moved out of the bottom of the screen and became full state screens with a clear action (go back and try again, or contact the payee). Camera permission denied got a different treatment: a bottom sheet with a direct link to the device settings, since that's where the fix was.
New labels on dynamic QRs: "Juros" became "Juros por atraso" and "Multa" became "Multa por atraso." Every field unique to dynamic QRs was marked as conditional in the spec.
Review for fixed-amount QRs: a review screen variation for QRs where the payee sets the amount and the payer can't change it. The edit option disappears in this case, and a visual indicator makes the lock explicit.
Edit amount: a separate sub-flow for static QRs with no fixed amount. People can change the amount before confirming, and they return to the review after editing.
Bottom sheet: confirmed as the right pattern for confirmations mid-flow, instead of interrupting with a full page. Used the same way in the QR and copy-paste flows.
High-value flow changes
Biometrics in the right place (individual accounts): biometric authentication moved to after the review, not before, so people see what they're authorizing before confirming with their fingerprint or face.
Date selection (individual accounts): the date picker only appears when the transfer would go over the daily limit. If the chosen date is outside the allowed range, the feedback is clear.
Business approval, queued state: the after-hours scenario stopped being a blocking error and became a queued state with an expected send time. The requester's home screen shows a persistent card with the pending transfer.
Business approval, approver screen: it shows the full transfer details, who requested it and two clear actions: approve (green) and reject (red), with a required reason when rejecting.
Hi-fi design
The final hi-fi covered both fronts completely, following the tokens and component library of Inter's design system for iOS and Android.
QR code and PIX copy-paste
Main flows:
- Standard review, static QR: amount, payee and bank. The amount is editable when the QR type allows it.
- Dynamic QR with due date: review with due date, interest and late fee, shown based on what the QR contains.
- Dynamic QR with due date, interest and fees: the full dynamic QR with every conditional field active.
- Review for fixed-amount QR: a non-editable review for QRs with a preset amount.
- Review for fixed-amount QR with message: the same screen with an optional message field.
- Edit amount: sub-flow to change the amount before confirming.
- Bottom sheet: the confirmation and sub-action patterns used across QR and copy-paste.
- PIX copy-paste: the full flow, from pasting to review to confirmation.
Scanning errors:
- Camera permission denied (with a link to device settings)
- Expired QR (full error screen)
- Invalid or unreadable QR (full error screen)
- Network error mid-flow
High-value PIX
Individual accounts:
- Limit check (how much of the PIX limit is used and how much is left)
- Biometrics (after the review)
- Date selection (only when the limit is exceeded)
- Confirmation with date and amount
Business accounts with approval:
- Amount to pay / set amount: amount and payee, picked from the company's contact list
- Review: full transfer review for business accounts (company tax ID and legal names)
- Business hours: after-hours queued state with an expected send time
- Approval: the approver's screen, with transfer details, approve or reject, and a reason when rejecting
Handoff & design QA
At Inter, handoff was a team process. Finished flows went into a shared handoff section of the design file, annotated and linked to a spec document. I walked through my flows in a session with the front-end squad responsible for PIX and in a second session with the QA team.
Design QA had its own round, beyond my own review. A QA analyst tested the implementation against the spec and wrote a formal report. I went through the findings with the product manager and prioritized fixes before the release candidate was approved.
What QA found:
- Dynamic QR review: interest and fee fields showing even when the QR had no such data, because the filtering logic wasn't implemented correctly
- Business approval: the approver's notification opened the app's home instead of the pending approval screen
- Business after hours: the expected send time on the queue showed UTC instead of Brasília time
- QR errors: the camera-denied state appeared before the system had even asked for permission, creating a false error
Everything was fixed before PIX launched.
Reflections: then and now
I designed these flows as a mid-level UX designer. Looking back as a senior, there are things I'd do differently. The result wasn't wrong; what changed is how I work and what I prioritize.
In the process:
Back then I treated the brief as the absolute truth and moved straight into execution. Today I'd question its assumptions earlier, especially in the business approval flow. The business hours rule was a product decision that came from compliance, but it made the experience worse, with requests sitting in a queue. As a senior I'd raise that sooner and suggest alternatives, like approval through push notifications after hours or a different time window. The constraint may well have been real, but design should stand up for users at the brief stage, not only work around the rules during execution.
In testing:
Lo-fi tests followed the research team's schedule, not the risk of each flow. Today I'd push to focus testing on the moments that demand the most trust: the review screens, where irreversible decisions happen, and the error states, where people lose confidence in the product. We tested entire flows, and the focus got diluted.
In cross-team alignment:
In a 50-person design team, there were dependencies I didn't always flag in time. The PIX flows touched the design system team (component behavior), the notifications team (the approver's push content) and the home screen team (the pending transfer card). I logged these as handoff items instead of tracking them as design dependencies. Today I'd treat those touchpoints as part of the flow, not as footnotes.
What still holds up:
The error hierarchy in the QR flows aged well. Splitting denied camera, expired QR and invalid QR into separate states, instead of one generic error, was the right call, and testing confirmed it. Treating errors as part of the design from the start, not as edge cases to polish later, is something I've taken into every project since.