Improving Help & Support for OneCard

Problem

Users rely on customer support during high-stress situations like fraud, failed payments, or blocked cards.

However:

• Help was fragmented across multiple flows

• Chatbot interactions felt robotic and untrustworthy

• Users depended heavily on human agents even for solvable issues

👉 Result: Slow resolution, high anxiety, and poor trust in the system

Key Insight

Not all queries are equal.

From research, user queries clearly split into:

• 60% Emergency queries → high urgency, panic-driven

• 40% General queries → low urgency, informational

👉 Existing system treated both the same → this was the core problem

Supporting Insights

• Users are willing to use chatbots, but don’t fully trust them for critical issues

• During panic, users want fast, clear, and direct actions

• Users prefer self-service, but only if it feels reliable

• Lack of empathy in chatbot → breaks trust instantly

Design Direction

Based on insights, I defined 3 core principles:

1. Prioritize urgency over structure

Emergency actions should be immediate, not buried

2. Reduce dependency on support

Enable users to solve issues within the product

3. Build trust through interaction

Make the system feel responsive, guided, and human

What I Changed

Instead of only improving the chatbot, I redesigned the entire Help experience.

1. Unified Help Experience

Problem

• “Need Help” and “Quick Actions” were separate

• Users had to decide where to go during panic

Solution

• Merged both into a single “Need Help?” entry point

• Prioritized emergency actions at the top

👉 Users no longer think, just act.

2. Smarter, Contextual Chatbot

Problem

• Robotic responses

• Users forced to leave chat for actions

• No emotional awareness

Solution

• More empathetic and guided responses

• Enabled in-chat actions (block card, select transaction, etc.)

• Smooth fallback to human agent when needed

👉 Chatbot becomes a problem-solving interface, not just support

3. Simplified “Block Card” Flow

Problem

• Too many controls → high cognitive load during panic

Solution

• Introduced “Swipe to Block”

• Reduced decision-making to one clear action

👉 Designed for speed under stress

4. Proactive “Set Reminder” Feature

Problem

• Users forget bill payments → avoidable stress

Solution

• Introduced 1-tap reminder setup

• Auto-configured based on billing cycle

👉 Shift from reactive → proactive support

End-to-End Experience

The redesigned system:

• Handles urgent issues instantly

• Supports self-service resolution

• Builds trust through better interactions

Expected Impact

(Hypothetical, based on design decisions - can be validated post-implementation)

• Reduced time to resolve critical issues

• Reduced dependency on customer support agents

• Increased trust in chatbot for high-stakes scenarios

• Increased user confidence during emergency situations

Key Learnings

• Designing for emotions (panic, anxiety) changes everything

• Chatbots shouldn’t just respond - they should guide decisions

• Real impact comes from system-level thinking, not isolated features