CASEE (Virtual Helpdesk Assistant)

Project began with the goal of setting up a virtual agent in 3 months that could resolve the easy "how-to" help desk calls (30% of total) on 8 used cases with a deflection rate of 58%. Currently, this virtual agent has grown to now handle 1000 conversations per month, with 1020 user intents and 48% of requests solved without escalating to a human specialist.

Problem definition

Deflect 5-10% of “easiest” help desk calls while maintaining 4/5 CSAT.

User journeys

We mapped 8 initial "how-to" use cases that the agent was designed to resolve.

One of 8 use cases identified for the pilot

Design iterations

The avatar, "CASEE" (Caring And Smart Engineered Entity was designed to appeal to an international audience and to avoid gender stereotypes. The brand identity was defined as caring and efficient. I set up a workshop to get stakeholder alignment on brand identity and avatar personality. These were then tested with focus groups and individual users over several rounds of development.

Early user testing

One-sheeter of final design

Integration within website

I wired framed out the integration of the chatbot with our existing user-facing knowledge objects.

MVP interface