Lucy is an integrated data application platform for the American Express ecosystem, providing developers with data and application abstractions to simplify product development effort while satisfying complex enterprise requirements.
Data Application
Visa intelligence commerce
Visa Intelligent Commerce (VIC) enabling agents, issuers, merchants and developers to transact with trust, confidence, speed, and intelligence.
(To comply with my non-disclosure agreement, some images presented in this project have been treated as unreadable on purpose. The information in this project does not necessarily reflect the view of Visa)
Overview
Design for trust, confident, speed, and intelligence.
Role: Lead product designer
Scope: Drive the end-to-end UX strategy to build consumer trust and confidence in AI-driven transactions.
Timeline: March 2026 - June 2026
Responsibility:
-
Led UX Strategy that focused on maximizing trust and user confidence during agentic transactions.
-
Partnered with cross-functional research teams to translate complex insights about AI shopping behaviors into actionable design strategies.
-
Defined Agentic Use Cases: Converted consumer insights into agentic commerce features, defining how users interact with transactional AI.
-
Cross-functional collaboration: Transformed complex business goals, partner requirements, and consumer data into intuitive, high-fidelity design solutions.
-
Global localization: Collaborated with regional design teams to adapt and scale the core user experience across diverse regional markets.
-
Cross-department handoff: Provided high-quality design assets to marketing and sales teams to ensure unified product storytelling.
Challenge
As AI agents begin to reshape digital commerce, partners are looking to Visa for leadership, infrastructure, and strategic guidance.
Visa Intelligent Commerce (VIC) is Visa’s portfolio for the agentic era, enabling AI agents, issuers, merchants, and developers to transact with greater trust, speed, and intelligence.
However, adoption remains uncertain.
Consumers and businesses are hesitant to let AI agents initiate payments, while issuers, acquirers, merchants, and payment processors must navigate new questions around identity, authorization, security, accountability, and user control.
Our mission was to identify the experiences, capabilities, and trust signals required to make agent-enabled commerce understandable, trustworthy, and valuable for every participant in the payment ecosystem
How might Visa build the trust and secure infrastructure needed for consumers and businesses to confidently adopt agent-enabled commerce?
Establish Trust and Control in Agent-Initiated Payments
-
Enable secure payment authentication for agent-initiated transactions
-
Clarify accountability and dispute resolution when an agent makes an incorrect or unauthorized purchase
-
Help merchants identify and accept trusted purchasing agents
-
Provide transparency across the full transaction journey
-
Enable interoperability across agents, merchants, issuers, wallets, and payment networks
Early Impact
-
Secured buy-in from key business partners
-
Showcased the concept to clients at the Visa Payments Forum
Prototype Demo with Claude
The Design Journey
Understand the Space
Research to Identify Key Elements to Drive Adoptions
Back in 2025, Visa had announced Visa Intelligent Commerce as the groundbreaking initiative set to transfer how we shop and pay. Given AI shopping will introduce new shopping behavior into consumers' daily life, we decided to test some agentic prototypes globally.
We partnered with regional design team, UX research team, marketing research and analysis, and product to launch 3 key research: a globally quant study, Quality test designed for authentication (VPP), Diary study
Below are selected sample flow we tested in 4 main regions:
Sample flow for US and Singapore
Sample flow for UK
Sample flow for Brazil
(Above are some selected pages from the research report. Credit to: Loan Tran, Daniela, Yan, Ellie)
1. Understand the Space Globally: Agentic Commerce Payment Stage
(Above are some selected pages from the research report. Research report credit to: Jack and Adam)
The VIC quantitative study survey 4,000 consumers across US, UK, Brazil, and Singapore.
2. Deep Dive into the Behavior: Consumer Diary Study
The study tracked a five-day journey from initial discovery to purchase execution. The goal is to learn how consumer use agentic AI tools for shopping, what excites them and what are common barriers or concerns, and identify key growth opportunities for Visa.
-
Frustration occurs when transactions can't be completed within the AI tool
-
Concerns about over-automation making some hesitant to share their payment info
-
Display accurate product prices upfront was expected, but AI failed to do so
3. Deep Dive into Authentication: Agentic + Visa payment passkey
We partner with Visa payment authentication team to launch qualitative study to understand: consumers' expectation of agentic commerce, reaction to purchase from multiple-merchant and intent-based agentic transactions, interest in using Visa payment passkey, and the impact of integrating VPP into agentic transaction.
(Above are some selected pages from the research report. Research report credit to: Yan and Ipso researchers)
Key Learnings
Convenience is the core appeal
Trust is the adoption bottleneck
Merchant Transparency is a major gap
Visa has a strategic Trust Advantage in the payment-stage moments
How to Build Trust in Agent-Initiated Payments
There were time period working I found myself lost in all research findings and countless brainstorming and "working sections" with AI tools like ChatGPT, Gemini, and Claude. Eventually, I decide to take a step back and start with understanding the human behaviors and motivations.
Questions that I'm starting to ask myself are:
-
What makes people trust someone or something?
-
How does people build trust when it comes to payment?
-
What is the differences between trusting agent and trusting Visa?
-
How can we expand people's trust in Visa to agentic commerce?
What inceases anxiety and fear?
Black box decisioning
Fully antonomose
Uncertainty
Loss of control
Anxiety
Fear
How to design for reducing anxiety/fear while increasing trust/confidence?
Fully autonomy
Human in the loop
-
Provide option to choose when approval is needed
-
Biometric authentication
Unfamilarity
Familiar pattern
-
Leverage Visa brand
-
Leveraging existing online shopping behaviors
Blackbox decisioning
Clear communication
-
Merchant transparency
-
Adding tax and shipping cost
Unknown risk
Error handling
-
Provide order editing
-
Clear dispute process
-
Edit order with agent
"Consumer should not need to trust the AI agent itself but the system governing it"
"Payment trust isn't built by features. It's build by reducing uncertainty at every step."
Rapid Interactive Prototyping
Prototyping with Claude Code and Figma Make
Although some early design ideas are developed in Figma, we quickly switch to Figma make and Claude Code to explore prototyping with AI tools.
(Snapshot of some early prototyping results)
(Snapshot of error result)
With the assistant of Claude code, I was able to quickly turn the Figma design into a working prototype.
(Snapshot of some early prototyping results)
In the age of AI, it is easy to build something quick and looks nice. This also leads to an increase of decision making cost. Having many "quick and look nice" design assets does not always leads to a solid product design.
“We are moving into a place where the cost of execution goes way down. Then, everything around it value goes up. The two thing that is going to be really really hard is figuring out what should we build and is it good enough”
Developing the Key Moments
Designing Building Block
Provide Clear Information for Tax and Shipping
From one of the key research, we have learned that consumer get frustrated when the tax and shipping fee is unclear("Displaying accurate product price upfront was expected, but AI failed to do so"). Therefore, when user set up their desired price, we design the AI to ask for clarification when user set up their desired price. If they set desired price as 500, AI will follow up with questions if it includes tax and shipping or it is only for item prices.
(Sample flow: Reflects work-in-progress designs and concepts at the time of case study creation, not the final product.)
Designing Building Block
Human-in-the-loop VS Autonomous
Provide consumer with clear options to choose how much involvement they want in the checkout process.
1. Human in the loop: Agent makes purchase after user confirmation (authentication at time of purchase)
2. Automated: Agent makes purchase with authorized purchase intent (authorization prior to transaction)
Designing Building Block
Trust Signal during Checkout Flow
Utilizing familiar UX patterns, such as the shopping cart and checkout confirmation, provides consumers with a sense of comfort and reassurance. Additionally, leverage brand trust (Visa) and biometric authentication to signal security transaction.
(Sample flow: Reflects work-in-progress designs and concepts at the time of case study creation, not the final product.)
Collaborate with cross-functional stakeholders to validate designs and engineer new capabilities.
Post-purchase order modification
Deploy an AI assistant to handle post-purchase order editing and cancellation requests through conversational interfaces (e.g., chat, voice) within a reasonable, designated time window to streamline user experience and eliminate manual support overhead.
Leverage agent automation for low-engagement commerce
Target high-value use cases centered on routine, friction-heavy buying behaviors. Our UX study confirms that consumers eagerly delegated purchasing control for recurring payments, repeated household goods, and predictable commodity shopping.
Evaluate agent performance via live User Testing
Conduct structured live testing sessions with target consumer groups to evaluate the VIC agent, and gather behavioral data on user interactions, task completion speed, and consumer trust levels.