This work was part of a broader Service Centre transformation at Flybuys, focused on reducing operational costs, centralising channels, and improving service efficiency. A key initiative within this transformation was replacing a legacy, static chatbot with a scalable, fully integrated chat experience built on Amazon Connect.

Chat was one of the first digital channels integrated into Amazon Connect, laying the foundation for consistent, connected support experiences across voice and digital channels and opening the door to future automation and AI-driven self-service.

Flybuys’ existing chatbot was limited in capability and disconnected from core service systems. Agents relied on multiple tools to manage member interactions, frequently switching between platforms to resolve even simple requests.

At the same time, data showed that a large proportion of member contacts related to simple, repeatable tasks, such as checking points balances, updating personal details or understanding account status. These interactions were costly to handle via voice and unnecessarily complex for both members and agents.

The challenge was twofold:

• Improve the member experience by enabling faster, more confident self-service
• Create a low-cost, scalable channel that could reduce agent dependency while remaining available 24/7

I led the UX and experience design across the chatbot initiative, working closely with product managers, principal engineers, and developers throughout delivery.

My responsibilities included:

• End-to-end experience design for members and agents
• Defining chat flows
• Creating reusable components and detailed build annotations
• Supporting QA with a focus on usability and accessibility

Timeline: 12 weeks

Understanding the user

To ensure the design was grounded in real member behaviour, I conducted extensive research into both member and agent experiences across the chat platform. This included analysing usage data, escalation patterns, common queries, and operational processes.

Key insights from this research included:

Members:

• Repeated identity verification
• Unclear wait times and session timeouts, causing friction and frustration
• Incomplete or missing context for self-service

Agents:

• High volume of low complexity enquiries
• Fragmented tools and systems
• Incomplete or missing context
• The need to repeat verification slowed resolution
• Chat operated in isolation from voice and other channels

I explored multiple pathways to improve outcomes incrementally while staying within scope and platform constraints.

Exploration 1 - Enabling self-service through content

I explored multiple pathways to improve outcomes incrementally while staying within scope and platform constraints.

Ideally, members would receive answers directly within the chat interface dynamically tailored to their question, device, and context. However, AI-driven FAQ responses were not yet in scope for this release.

Design decision:

For this phase, members are redirected to the FAQ page when appropriate, while ensuring the chat experience sets expectations clearly and remains cohesive. This created a bridge solution that delivers value now while preparing for AI-powered responses in future releases.

Exploration 2 – Identity

One of the most impactful explorations focused on identity verification within chat.

I explored several approaches, including:

• Prompting members to sign in early using existing Flybuys credentials
• Capturing personal details step-by-step and verifying via Amazon Connect
• Using structured forms to streamline verification

Design decision:

Early sign-in was the most convenient approach, allowing members to use existing sessions, especially in the app where they were already signed in. This also allowed agents to skip identity verification entirely once the member was authenticated.

As a result, friction was significantly reduced for both members and agents, directly contributing to lower average handling times.

Exploration 3 - Designing for continuity

Another issue surfaced around session management. Members often stepped away from chat and returned later — only to find their conversation lost.

To address this, I designed a solution where:

• Chat sessions remain active for up to 48 hours when escalated to an agent
• Agents can explicitly close conversations once confirmation is received
• A post-contact survey is triggered at the appropriate moment
• This reinforced a sense of continuity and respect for the member’s time.

Once the core flows were defined, I moved into detailed design and delivery:

• Created reusable components aligned to Amazon Connect capabilities
• Documented edge cases, states, and accessibility requirements
• Partnered closely with engineering to ensure feasibility
• Updated design system elements to reflect Flybuys’ refreshed brand

This level of detail ensured the experience scaled cleanly and could support future enhancements without rework.

In September, the legacy service centre management tool was successfully decommissioned and replaced with Amazon Connect.

The chatbot MVP launched with a solid foundation and delivered measurable impact:

• 3% reduction in average handling time
• 49% reduction in technical costs
• Seamless escalation, with agents receiving full conversation context
• A scalable base ready for AI-driven self-service in future phases

More importantly, it shifted chat from a fragile experiment into a credible, strategic service channel.