Turning Routine Queries into Self-Serve Success
Exploring how conversational AI could ease call centre pressure and give LV= staff more time to focus on what matters.
The problem: LV=’s call centre teams were spending a disproportionate amount of time on routine tasks — password resets, customer verification, and directing people to simple information. This reduced their capacity to deal with complex or high-value queries and impacted customer experience.
The solution: We set out to design a chatbot proof of concept (POC) that could:
Collect key customer details before an agent joined the conversation.
Direct users to relevant self-serve information on the website.
Free up call centre staff to focus on more important issues.
The chatbot was designed to integrate with LV=’s existing live chat, demonstrating how conversational AI could become a scalable part of their support ecosystem.
Skills used: Conversational UI design | Workshops & ideation | Wireframing & prototyping | Usability testing | Developer briefing | Copywriting | UI design
Process
Initial flows
Working with the consultant team, I mapped out full conversation flows — breaking them into structured sections like introduction, intent capture, and resolution. These were reviewed with LV= stakeholders to prioritise the types of queries the POC should handle.
Conversation flows
The flows were then refined into visual conversation maps showing how each intent would branch and connect. These became the backbone for building and training the chatbot in Salesforce Einstein.
Agent’s screen
Alongside the bot, I collaborated with the developer on redesigning the agent interface. By shadowing call centre staff and reviewing their own sketches, we created a prototype layout that surfaced essential customer information upfront while providing easy access to a knowledge base for deeper support.
In build
I worked closely with the developer to continuously test the chatbot, ensuring smooth handovers and no conversational dead-ends.
Usability testing
We ran usability sessions at LV=’s Bournemouth office with call centre agents. Testing the prototype in context provided valuable feedback, which we distilled into clear recommendations for future iterations.
Outcome
The POC proved that a chatbot could handle routine queries effectively, cut down call centre workload, and support staff with smarter tools during calls. It gave LV= confidence in pursuing conversational AI as a viable long-term solution for customer support.
