FXpress
AI-Augmented Mobile App for Cross-Border Remittance
Sending money internationally shouldn't require guesswork. FXpress is a conceptual mobile banking app that predicts the best time to convert currency — and asks before it acts.
Client
Conceptual · Fintech
Year
2025
Platform
iOS
Timeline
10 weeks
Context
Remittance is not a neutral transaction. For most users, it is a recurring act of care — sending money home to family who depend on it, every month, without fail. The anxiety surrounding it is not just about fees or exchange rates. It is about reliability, legibility, and the confidence that the money will arrive exactly as intended.
The market is crowded — Wise, Remitly, Revolut, Western Union all compete on speed and cost. FXpress enters at a different intersection: not just cheaper transfers, but intelligent ones. The core design challenge was not building a better remittance app. It was building one that earns trust in a context where users have historically had it exploited.
The Problem
People who send money internationally absorb fees they don't fully understand, guess at whether today is a good day to convert currency, and have no tools to act confidently rather than reactively.
Three things make this problem hard to solve:
FX timing is a universal pain point — but in different forms. Some users don't know what a good rate looks like. Others don't have time to track it. Others monitor it manually and still miss the optimal moment. The pain is not lack of access to rate information — it is the cognitive burden of having to act on it correctly, under time pressure, while managing other priorities.
The most valuable features are the most likely to generate resistance. AI-timed conversions, stablecoin routing, smart automation — these are exactly the features users are most skeptical of. The design challenge was not whether to include them, but how to make them feel safe, optional, and legible to each type of user.
The recipient side is non-negotiable. If the person receiving money has to learn something new, the sender will not switch apps. The product's complexity must never appear on the receiving end.
Research
Who I spoke to and what I learned
Four participants were interviewed — a marketing coordinator sending to Manila, a software engineer sending to South Korea, a warehouse supervisor sending to El Salvador, and a freelance UX designer sending to LATAM and Africa. Each represented a distinct point on the spectrum from manual-only to fully programmable automation.
The four findings that shaped everything:
1. FX timing frustration is universal. Every participant described the same problem in different forms — guessing at rates, Googling manually, missing optimal windows. The cognitive burden of acting at the right time, while managing other priorities, was the consistent pain point.
2. All four will accept crypto rails — if invisible. No participant refused crypto outright. What all four rejected was crypto as a user experience. The consensus was tolerance for crypto as infrastructure, with zero expectation of encountering it.
3. Outcome clarity is non-negotiable. All four participants, unprompted, described the same core need: before anything is sent, they want to see exactly what the recipient will receive. Not an estimate. The exact amount, fee, delivery time — before confirmation.
4. Trust is built through control, not features. Every participant framed trust in terms of control — the ability to see what's happening, approve before it happens, and stop it if needed. No participant framed trust as a function of brand or design polish alone.
"I'd need to see what the rate is, what my family's getting, and how long it'll take."
Perry — Warehouse Supervisor, Dallas TX
"I don't have time to track that constantly."
Jenni — Marketing Coordinator, Los Angeles CA
Meet Fexi
Fexi is FXpress's AI co-pilot — a recommendation layer, not an automation engine.
The distinction matters. Every research participant expressed conditional openness to AI assistance — they would use it if they could see the reasoning, understand what was being recommended, and decline without friction. None of them wanted automation they did not sanction.
Fexi is designed around that condition. It monitors market rates, identifies optimal conversion windows, and surfaces a recommendation with clear reasoning — projected savings, rate trend, confidence level. It then waits. Nothing happens until the user says yes.
This is not a trust-softening measure. It is an accurate representation of what AI assistance should be in a high-stakes financial context: a co-pilot that explains its reasoning and always waits for a yes.
Key Design Decisions
Decision 1 — One default experience, one progressive layer
Research revealed four distinct users with four distinct automation tolerances — from Perry's "I always click" to Sadi's programmable rules. A single product had to serve all four without forcing any of them into a flow that felt wrong.
The solution: a conservative default that works for Perry and Jenni — one confirmation, no jargon, full outcome preview — with a progressive power layer that surfaces for Kang and Sadi. Rate thresholds, routing visibility, action logs, programmable triggers. Same product, adaptive depth.
Decision 2 — Crypto is a backend decision, not a frontend one
All four participants expressed tolerance for stablecoin routing — provided it was completely invisible. The word "crypto" reads as risk for the majority of FXpress users. The design requirement was abstraction — zero crypto terminology in the default flow, with optional visibility for users who want it.
Kang's framing was the clearest signal: "My mom doesn't need to know what USDC is — as long as she gets her KRW."
Decision 3 — The confirmation screen is the most important screen
Outcome clarity was the most consistently cited trust requirement across all four participants. Before any transfer executes, the user sees: amount sent, fee, exchange rate, recipient amount in local currency, estimated delivery time. Nothing is approximate. Nothing is hidden.
This screen is not a legal formality — it is the moment the product earns trust or loses it.
Outcomes
FXpress is a validated conceptual design with a working prototype. The research confirmed the product concept is viable — all four participants said they would use it if it solved the timing problem without introducing new risk or confusion.
The automation spectrum insight — from Perry's full manual to Sadi's programmable rules — directly shaped the default/progressive layer architecture and prevented a common design mistake: building for the power user and leaving everyone else behind.
Key validated design principles:
Fexi recommends, user decides, always
Crypto abstracted by default, surfaced on request
Confirmation screen shows exact amounts, no estimates
Automation earned over time, not assumed from onboarding
Reflection
The research was stronger than I expected going in. The automation spectrum — four participants mapping cleanly across a range from manual-only to programmable — gave the design clearer direction than a single target user would have. That spectrum became the architecture.
What I would do differently: test Fexi's recommendation UI earlier and in isolation. The co-pilot framing — recommendation with approval, not automation — was validated conceptually through interviews, but earlier usability testing of the actual Fexi interaction pattern would have sharpened the UI decisions around how much reasoning to show, and when.
View Prototype
Previous
Next