Celero Xpress — Digital Banking

Overview

One platform. 30+ credit unions. 1.5M members.

Canada's credit union system wanted one digital banking experience that members would actually enjoy using. The long-term plan was to design web and mobile banking, suggest new features, and bring everything together into one experience for the brand, covering both personal and small-business banking.

I led UX across responsive web, iOS and Android. I also built and looked after Celero's shared, white-labelled design system, so every credit union could ship something consistent and accessible under its own brand.

What it led to

Xpress digital banking was rolled out by 30+ credit unions across Canada.
1.5M+ members saw the new design through usability testing and rollout.
Pilot groups showed a 20% increase in digital adoption and a 15% lift in feature usage.

Why this mattered

Banking had moved to phones. The old experience hadn't.

Most members were checking balances and moving money on their phones, but the existing online banking was built for a desktop screen and a different decade. Simple tasks took too many taps, useful features were buried in menus, and members kept calling their credit union for help with things the app should have made obvious.

For the credit unions, that wasn't just a design problem. Every confused member meant a support call, and every member who gave up on the app was a little more likely to move to a big bank with a slicker one. Xpress needed to feel modern and trustworthy, and still work for 30+ credit unions, each with its own brand and members.

The challenges

What made this hard.

One product, 30+ brands

Every credit union wanted its own colours and logo. The design had to look right for all of them without turning into 30 apps to maintain.

Security vs. speed

Members wanted to get in fast. Security needed one-time codes and extra checks. Both had to feel reasonable, not like hurdles.

Three platforms at once

Web, iOS and Android each have their own habits. Members switch between them, so it had to feel familiar everywhere.

A very wide range of members

Students to retirees, daily users to people who log in twice a year. Clear had to mean clear for all of them.

Money is unforgiving

A confusing transfer or a rejected cheque photo costs people real money and stress. Mistakes needed to be rare and easy to recover from.

Accessible in every brand

Every brand colour combination still had to meet WCAG, so the system had to stop credit unions from accidentally breaking contrast.

Try it

Use the Xpress app.

This is a working version of the Android app. Every tab and button does something, and your balances update as you go.

Sign in with the fingerprint button, or any password followed by any 6-digit code. Then try paying a bill, sending an e-Transfer, moving money between accounts, or depositing a cheque. You can also search your activity and change settings under More. If you enter a cheque amount other than $1,146.99, the app warns you it doesn't match the cheque. Use "Restart prototype" under the phone to start over.

How I used AI

Turning static screens into a working app.

The original Xpress designs were static screens. To show how the app actually behaves, I used AI as a design partner to rebuild it as the working prototype above. I set the direction and made the calls; AI did the heavy lifting on code and drafts, and I reviewed every screen against the original designs.

From screens to a clickable app

I described each flow, from sign-in and the one-time code to transfers, bill payments and cheque deposit, and AI turned the original screens into working code. Building it this way took days, where hand-linking static screens would have taken weeks.

Real words for every "Label"

The design files had placeholder buttons. AI drafted plain-language wording for each one, plus error messages like "That's more than the available balance", and I edited them to match the Xpress tone.

Made-up members and money

AI generated believable accounts, balances, payees and transaction history, so the prototype feels real without using any real member data.

Listing the unhappy paths

I asked AI to list everything that could go wrong: an empty password, a wrong code, an amount over the balance, a cheque amount that doesn't match the photo. Then I made sure each one has a clear, kind response in the app.

Android conventions

AI helped me check the prototype against Android patterns like back arrows, bottom sheets and Material dialogs, so it behaves the way Android users expect.

Testing and fixing

Each round, AI clicked through every flow, reported what broke, like a password field that wouldn't take typing or icons that didn't show, and fixed it. I tested the result before moving on.

The judgement stayed with me: what to build, how it should feel, and whether it was right. AI made it possible to get there in a fraction of the time.

Decisions I owned

What I decided, what I traded off, and who agreed.

The process above is how the work got done. These are the calls that shaped it, all agreed with the credit union stakeholders.

The tension: their brand vs. one system

Every credit union wanted its own branding and its own features. I kept them on one unified design system, built with our vendor ebankit, so each could show its own brand while the screens and flows stayed shared. That's what made it possible to roll out to 30+ credit unions instead of maintaining 30 separate apps.

The decision I pushed through

I argued for one design system over separate builds, and then helped each credit union through its rollout so the shared system actually got adopted, not just approved.

The constraint: accessibility compliance

Every credit union had to meet accessibility standards. I led the accessibility work on Xpress: I designed accessible UI into the shared components and added screen reader support across the app, so each credit union got a compliant experience through the same system.

Why I made these calls

The reasoning behind the big decisions.

Most of the work above came down to a handful of choices. Here's what I chose and why.

Put everyday tasks up front

Research showed members open the app to do four things: pay a bill, transfer, send an e-Transfer or deposit a cheque. Hiding those in a menu added taps to every visit, so they went straight onto the dashboard.

One system, not 30 apps

Thirty credit unions each wanted their own brand. Separate designs would have drifted apart and been impossible to maintain. A white-labelled system with swappable colour and logo gave each one its look while the screens stayed the same.

Explain security, don't just add it

Members saw one-time codes as the bank getting in the way. Saying why the code was needed, and falling back calmly when fingerprint failed, turned a frustrating step into one that felt protective.

Design the unhappy paths early

Support calls came from blurry cheque photos, wrong amounts, empty lists and lost connections, not from the main flow. Designing those moments first meant fewer members got stuck and fewer called in.

Review before money moves

A mistake with money is stressful and slow to undo. A clear review screen before every transfer and deposit costs one tap, and it lets members catch errors while they can still fix them.

Test with real members, early

Changes are cheap when they're still sketches. Running 40+ interviews and usability sessions while the design was still forming meant problems got fixed before they were built.

Looking back

How AI could have helped us ship faster.

When I worked on Xpress, AI wasn't part of our design workflow yet. Looking back, a lot of the slow, repetitive work is exactly where it helps most. Banking data is sensitive, so the rule would be simple: no real member data goes in. Made-up data, design files and public guidelines are fine.

30 brand themes in an afternoon. Generate each credit union's theme from its logo and colours, and flag any combination that fails WCAG contrast before a designer opens the file.
Faster research synthesis. With names removed, interview notes can be grouped into themes in hours instead of days, leaving more time to talk to members.
Realistic test data. Fake accounts, balances, payees and transaction history for prototypes and usability tests, so sessions feel real without touching real accounts.
Every edge case, written down. AI is good at listing unhappy paths like expired codes, partial cheque photos and duplicate payees, which gives the team a checklist to design against.
Clearer microcopy. First drafts of error messages and help text in plain language, then reviewed by compliance and the team.
Working prototypes in days. Instead of linking static screens, we could test a clickable prototype almost straight away. The prototype on this page was built that way, from the original screens.

None of this replaces talking to members or the judgement calls a designer makes. It clears away the busywork so more time goes into the parts that need a person.

Looking ahead

How AI can work in a regulated financial institution.

Banks and credit unions answer to regulators, auditors and their members, so they can't adopt AI the way a startup might. The opportunity is real, but only if every use is explainable, reviewed by a person and safe with member data. These are the places I see it helping most, and the guardrails each one needs.

Member help that knows when to hand off

An assistant can answer everyday questions like "When will my deposit clear?" or "How do I reset my password?" around the clock. It should say clearly that it's AI, never give financial advice, and pass the conversation to a person, with the context attached, the moment it's unsure or the topic is sensitive.

Spotting fraud and unusual activity

AI can flag patterns a person would miss, like a new payee followed by a large e-Transfer at 2 am. The design job is the alert: tell the member what looked unusual, let them confirm or block it in one tap, and keep a person in charge of freezing accounts.

Plain-language explanations

Fees, holds, term deposit rates and loan terms are full of jargon. AI can explain them in plain words on demand, as long as the wording comes from approved content and the official terms are always one tap away.

Faster compliance reviews

Before a release, AI can check screens and copy against required disclosures, accessibility rules and the brand guide, and list what's missing. Compliance still signs off, but they start from a checklist instead of a blank page.

Smarter document intake

Reading a cheque, a pay stub or ID and pre-filling the form saves members typing and cuts errors. The member reviews every pre-filled field, and anything the AI isn't confident about is flagged rather than guessed.

Insight from support and feedback

Thousands of call notes, chats and app reviews can be grouped into themes, with names removed, so product teams see what's frustrating members this week, not next quarter.

The guardrails that make it work

Member data stays protected. No personal information goes into public AI tools. Approved, private models only, with clear rules on what's stored and for how long.
A person makes the final call. AI can suggest, flag and draft. Decisions about money, credit and account access stay with people.
Every answer can be explained. If a member or an auditor asks why something happened, there's a clear, recorded reason.
Members always know. It's clear when they're talking to AI, and a person is always one tap away.
Checked for fairness. Models are tested so they don't treat members differently by age, language or background.

Designed this way, AI makes banking faster and clearer without losing the trust that credit unions are built on.

Celero Xpress: digital banking, reinvented.

One platform. 30+ credit unions. 1.5M members.

What it led to

Banking had moved to phones. The old experience hadn't.

What made this hard.

Problem, objective and outcome.

A clear, member-first process.

Listening to stakeholders and members.

Competitive analysis

Turning what we heard into people and journeys.

Deciding what mattered most.

The Xpress style guide.

Shipped on every platform.