Chat Support
Designing an AI-Powered
Assistant for Customs Brokers
AI UX
Conversational Design
Enterprise SaaS
Overview
Context
In the high-pressure world of customs brokerage, every minute counts. A delayed shipment, a missing tariff code, or a document upload error isn’t just an inconvenience — it’s a business risk. Customs brokers using the Sentinel Cloud ERP platform needed reliable, instant support to resolve critical issues within their regulatory workflows.
Yet their only options were email tickets, lengthy FAQs, and delayed phone support.
This wasn’t just a UX problem — it was an operational crisis.
We saw an opportunity to not just patch this gap, but to reimagine support through AI — crafting an assistant that’s helpful, trustworthy and human-centered, even when powered by algorithms.
The Challenge
Mission
Design an AI-powered chat support experience inside Sentinel Cloud ERP to:
Providing technical and regulatory support in real-time
Logging and managing communication exchanges
Facilitating job creation and job status tracking
Seamlessly escalating to human support when needed
Maintaining audit-compliant logs for every AI-assisted interaction
Key Goals
Cut resolution time by 40%
Increase AI self-resolved queries by 3×
Build operational trust in AI among risk-averse users in regulated industries
This wasn’t about adding AI for hype — it was about making AI invisible when it works and accountable when it matters.
Some notes goes here
My Role
Mission
As the Lead UX Designer, I owned the end-to-end experience strategy, AI-human interaction model design, and product alignment.
I collaborated with:
AI/ML Engineers
Product Manager and Support Lead
End Users (Customs Brokers and Ops Managers)
Seamlessly escalating to human support when needed
Maintaining audit-compliant logs for every AI-assisted interaction
From framing the problem space, defining AI UX principles, to prototyping, validating, and refining conversational flows — I led the experience design vision for this AI-powered assistant.
NDA Notice
Due to confidentiality with Unifo, full UI visuals and sensitive client data are omitted. However, the design process, decisions and outcomes represent my actual contributions. I’m happy to walk through anonymized flows and design logic in a live session.
The Discovery
Research & Insight
To build this responsibly, we immersed ourselves in users' pain points:
Key Insights
In a regulated workflow, users have zero patience for vague, opaque AI responses. They crave:
This shaped our foundational UX principle:
In high-stakes operations, AI should clarify, not mystify.
Design Process
We followed a structured, iterative process to redesign the experience without overwhelming users — especially those handling 50+ jobs per day.
The Solution
AI-Driven UX Strategy - Design Principles
AI Capabilities Integrated
Communication Logging: Automatically records job creation, API exchanges, and data conversions.
Job Creation Assistance: AI guides users through transport modes, customer details, and customs classifications.
Job Status Tracking: Instant status queries for submission acknowledgments, queries, amendments.
Document Management Guidance: Walks users through attaching and linking supporting docs.
Regulation Lookups & Error Troubleshooting: Contextual guidance on duty calculations and resolving system errors.
Collaboration & Iteration
Built this system through:
Weekly design working sessions with AI engineers
Continuous compliance review to meet regulatory standards
Issue mapping workshops with support managers
3 iterative prototype testing cycles with customs brokers
Prioritized user trust signals, fail-safes, and AI explainability toggles
The Impact — Result
NDA Note
Due to confidentiality with Unifo, full UI visuals and sensitive client data are omitted. However, the design process, decisions and outcomes represent my actual contributions. I’m happy to walk through anonymized flows and design logic in a live session.
Outcomes & Results
The redesign led to tangible improvements in user workflow, satisfaction and system efficiency.
Learnings
Here are my biggest Learnings
In regulated workflows, AI must clarify itself to earn trust.
Guided task-based conversations outperform open-ended AI chats.
Seamless escalation to humans isn’t a fallback — it’s essential UX infrastructure.
What’s Next?
Introduce AI voice commands for senior brokers
Visualize AI suggestion confidence scores for advanced users
Integrate micro-training tooltips during issue resolutions
Big Takeaway
Enterprise AI UX demands a balance of operational empathy, explainability, and practical AI integration.
