Industry
Design
At Algorand Agentic Commerce x402 Hackathon in Berlin, I teamed up with an AI engineer to build PackRoute Agent, an autonomous EU Food & Beverage packaging procurement agent, which aims to support European SMEs and enterprise procurement teams to speed up sourcing packaging suppliers with lower costs and the secured transaction.
Time
06 - 07 June 2026
Hackathon Organizer
Algorand
Tool
Cursor
Client
Algorand
My Role
UX & Product Designer
My Team
1 Product Designer, 1 AI Engineer
Impact
Faster decision on sourcing and procurement
Fewer manual PO errors
Lower cost in AI support for procurement by pay-per-request micro-payments and explainable recommendations
Context
During a 2-day Algorand Agnetic Commerce x402 hackathon, I teamed up with an AI engineer to design an agent-assisted procurement interface where users needed to trust AI recommendations in a regulated cross-border transaction. The core challenge: how do I design for autonomy while maintaining user confidence and transparency as designing for digital identity verification, where trust is non-negotiable.
Business Problem
Consumers / Users
European SMEs and procurement teams spend significant time and cost managing sourcing and purchasing through spreadsheets and email-based workflows.
Pain points:
No transparency into how suppliers are vetted
Users lack confidence in AI-driven decisions
Higher operational costs and increased procurement risk
Verification barriers slow adoption
Regulatory compliance is manual and error-prone
Business Objectives
Enable fast and trustworthy approval of AI actions
Reduce security risk while improving efficiency and human–AI collaboration with Algorand x402 micro-payment
Support seamless, transparent experiences for SMEs and enterprise procurement teams
AI Product Workflow= AI-assisted Design + Human Oversight
01 Brainstormed product ideas, defined user problem and product scope with Claude
02 Built prototype with Cursor
03 Iterated solution with Cursor & Claude Code
04 Created Pitch Slides & Demo video with Claude Design
Constraints
Time
Only 2 days to finish building prototype with Algorand integrated and creating pitch slides and DEMO videos
Knowledge
Both my teammate and I are new in FinTech and Blockchain.
Design Solution: PackRoute Agent
PackRoute Agent DEMO Video
We built PackRoute AI agent that sources cross-border packaging suppliers, pays for intelligence via x402 micropayments on Algorand, and completes supplier checkout within owner-defined spending rules.
Design Discovery & Design Decisions
Problem Statement
European SMEs and Procurement teams need to efficiently and effectively find the most suitable and reliable suppliers under regulatory compliance because they currently spend long time and high costs on supplier discovery, manual comparisons, and approvals and they don't trust that AI agents can help solve their problem.
Design Hypothesis
Hypothesis
Designing an AI agent which communicates users about the reliable sourcing context, decision support, and transparent blockchain transaction status will increase customer trust and conversion rates.
Design Changes
Designing simple identity verification user flow with progressive disclosure context for sourcing support and decision request and adding simple block chain transaction status dashboard for showing efficient, low-cost, secured, and transparent working process.
Success Metrics
10% increase in customer satisfaction ratings and 10% decrease in cost and time spent on procurement.
UX Design Decisions Throughout User Flow
For more efficient cooperation with AI engineer to complete prototype in time, I gave him advice on UX improvements and he iterated the prototype accordingly after the first version of prototype was built. When the prototype was ready, I used Claude Design to create pitch slides and demo video within 2 hours.
The user flow below shows design decisions we agreed on for the PackRoute Agent and the workflow of the Agent via Algorand x402 payment.
Scenario
Designing for transparent AI collaboration when the user opens the PackRoute agent dashboard:
Simple onboarding guidance by agent
Transaction status session card gives human oversight over full autonomy
Easy identity verification for the user using customized sourcing service and micro-payment:
Security/permissions handling without passowrd and API keys
Progressive-disclosure of critical procurement context inspired by regulatory best practices in identity verification that lets users understand why the agent recommends a supplier before confirming.
AI transaction workflow with Algorand in the user's agent wallet. PackRoute autonomously:
Pays $0.01 USDC (x402) for packaging supplier price index
Pays $0.05 USDC (x402) for cross-border freight quotes
Compares landed cost and selects best supplier
Pays $0.10 USDC (x402) to confirm supplier checkout
AI-assisted transaction workflow with human guardrails: Rather than fully autonomous execution, PackRoute requires users to approve spending limits upfront and validate the final supplier selection.
Pitching the PackRoute Project
Takeaways
01
Proactively communicating with the engineer about the working progress to give the helpful UX design advice on building the product with better UX.
02
Prioritizing simplifying the user flow to gain the user trust with AI agent decision and support
03
Understanding regulatory and security constraints is fundamental for trustworthy design.
Conclusion
Through building Agentic commerce product integrated with Algorand with AI engineer, I learned how to design the simple AI workflow for users in FinTech and Blockchain areas.
What I would do differently next time
Define a "Security & Trust Design System" and design principles upfrontfor communicating algorithmic decisions, handling permissions, showing compliance status
Conduct accessibility & compliance audits early by using WCAG and regulatory alignment as design inputs, not QA checkboxes
Validate user trust through testing, even in tight timelines (30-min user test for shipping faster)
Algorand Agentic Commerce x402 Hackathon Event Website on Luma
Prototype on Github
Pitch Slides on Github
Original Demo on Github