Empathetic AI for Sun Life: Designing empathetic conversations for an AI-powered customer experience.

Conversational AI During COVID-19: Designing an Empathetic Machine-Learning Chatbot

Role: Content & Behavioural Strategy Lead
Contribution: Conversation strategy + chatbot launch scripts + ongoing conversational optimization
Collaboration: Content, UX, AI/ML and governance
Focus: Improving question-and-answer matching, conversational flows and supporting pages

The Challenge
The Psychological Friction: During the pandemic, consumer anxiety surrounding illness and mortality spiked, leading to an unprecedented demand for life and health insurance. However, user testing revealed a profound emotional barrier: potential clients experienced intense feelings of vulnerability and shame when forced to explain their lack of financial safety nets to a live human advisor.

The Behavioural Shift: To avoid this emotional friction, a massive segment of users pivoted toward a self-service model, preferring to research and purchase complex insurance products entirely on their own.

The Corporate Constraint: Financial and insurance products are governed by strict regulatory frameworks. Every piece of automated dialogue had to balance immediate user empathy with ironclad legal transparency, ensuring zero liability or misrepresentation.

The Goal: Build and launch a direct-to-consumer conversational AI chatbot that provides real-time, self-directed answers, lowering the barrier to entry while upholding rigorous enterprise compliance standards.

BEHAVIOURAL INSIGHT → CONVERSATIONAL EXPERIENCE → CONTENT → UX → MACHINE LEARNING → ONGOING OPTIMIZATION

A behavioural-science approach informed not just the chatbot conversation, but the broader system around it, from content and UX to machine-learning feedback loops and continuous optimization.

My Strategic Process & Cross-Functional Governance
Deploying an unmonitored conversational interface within a highly regulated insurance environment requires immense operational oversight. As the strategy lead, I designed and managed a rigorous weekly content governance framework to bridge the gap between human emotion and corporate compliance:
The Triple-Stakeholder Review Matrix
Every single line of the automated conversation script was cross-examined in intensive, hour-long collaborative blocks with three distinct enterprise pillars:
    1. Legal & Compliance: Carefully auditing dialogue flows to ensure definitions of coverage, eligibility, and risk were legally sound, seamlessly embedding mandatory disclaimers without fracturing the user experience.
    2. Sun Life Brand: Ensuring the voice remained authoritative, reassuring, and aligned with enterprise standards, stripping out robotic transactional phrasing.
    3. UX Strategy: Championing the user’s cognitive load, dynamically adjusting content density to prevent information overload within the chat window.

The Machine-Learning Optimization Loop
The launch was only the beginning. Because this was a machine-learning initiative, the system relied on continuous data to train its natural language processing (NLP) models:
  • I established a recurring audit process to review failed intents (questions the chatbot couldn’t confidently map).
  • I iteratively re-authored and refined the conversational scripts based on actual real-world user queries, directly training the machine learning model to better understand intent variations, colloquialisms, and anxiety-driven questioning.

The Outcome
The Measure of Success: Launched and optimized throughout 2022, the chatbot successfully bypassed the advisor-shame barrier, resulting in increased user growth and adoption across the direct-to-consumer digital portfolio.

Enterprise Automation Scale: By proving that automated machine learning can safely navigate high-stakes insurance dialogues, this initiative successfully validated conversational AI as a highly scalable, trusted acquisition channel for Sun Life.