Chatbot UX

The goal was to implement a chatbot on the Choices platform landing page and internal flow to improve user engagement, answer common questions in real time, and increase conversions. The platform's purpose is to move users' savings into Sun Life to keep them growing. The previous design lacked immediate support, leading to a high drop-off rate caused by unanswered queries during the decision-making flow.

The work covered conversational flow design, chatbot model selection, response copy, widget placement, seamless live agent handoff, and a post-response feedback mechanism for continuous improvement.

Customer support volume clustered around a predictable set of repeatable intents, yet the existing experience handled them poorly. The absence of immediate support created three measurable failure modes.

Key Challenges

• Delayed response time, users had to wait for email responses, leading to frustration and abandonment.
• High bounce rate caused by unanswered queries during the decision-making flow.
• Low engagement on the landing page, users hovered around FAQs but left without taking action.

Research Insights

• User interviews found that users had repetitive questions about products, fees, transfer methods, and retirement policies.
• Users consistently hovered around FAQs but left without acting, passive information delivery wasn't enough.
• Competitor analysis revealed that chatbots significantly improved customer engagement and reduced drop-offs when designed with clear intent taxonomy.

The design thinking process structured exploration across four stages: Empathise, Define, Ideate, and Prototype, building from user research through to a validated hybrid chatbot model.

Empathise

User tests and feedback analysis identified the core pain points. Interview questions focused on what information users sought, what caused them to abandon, and what would have kept them engaged.

Chat Widget Placement

The widget was positioned in the bottom-right corner for consistent, low-friction access. This follows established conversational interface conventions, reducing the cognitive effort of finding support without competing with primary page actions.

The redesigned system solved the drop-off problem through architecture rather than content volume, every path through the bot led to a defined resolution state, even when that state was a graceful handoff to a live agent.

• Hybrid chatbot model combining rule-based accuracy for common intents with NLP flexibility for variation.
• Bottom-right widget placement for always-accessible support without disrupting primary page interactions.
• Quick action buttons surfacing top intents immediately on open, no typing required for the most common queries.
• Seamless live agent handoff with full conversation context passed automatically, reducing repeat explanation.
• Post-response feedback loop enabling continuous improvement of bot responses from real usage data.

After launching the chatbot, performance was tracked over 60 days, demonstrating improvements across all key indicators.

• A well-placed and intuitive chatbot significantly improves engagement, placement and discoverability are as important as response quality.
• Combining AI with human support ensures a seamless experience: the hybrid model outperformed both pure rule-based and pure AI approaches for this intent profile.
• Continuous monitoring helps refine chatbot interactions, the post-response feedback loop was the most valuable ongoing improvement mechanism post-launch.

Next Steps

• Optimise chatbot responses based on new user queries and intent patterns surfaced by the feedback loop.
• Extend the chatbot to the entire enrolment flow, not just the landing page.

The Chatbot UX demonstrates that conversational design is a trust problem before it is a flow problem. The highest-leverage decisions weren't about intent coverage or NLP accuracy, they were about what happens when the bot reaches its limit and whether users feel handled or helped.

Designing the handoff as a first-class experience, not a fallback, and building a feedback loop into the product from day one produced a system that improved itself after launch and maintained user trust even when it couldn't resolve an issue independently.