Envision Customer Service AI Agent Experience
My Role
Executing on design
Team
1 Designer Lead
2 Product Managers,
1 Engineering Lead,
Omni-Channel Team,
Customer Support Team,
Design System Team
Business Problem
TriNet's customer support team was overwhelmed with 35,000 live chat sessions monthly as users bypassed extensive knowledge articles to seek direct help.
This created a costly cycle: frustrated users waiting for answers while support staff handled repetitive inquiries that could be self-resolved.
Goal
An intelligent assistant that delivers personalized, contextual help, reducing support volume while improving user satisfaction.
Timeline
3 months, 2024
UX Outcome
Conceptualized and delivered solutions for current adaptive stage and ideal future state.
Produced mockups of key use cases, facilitated cross-team discussions, informed platform AI capability planning across product ecosystem.
Transform reactive customer service into proactive, self-service experiences through guided walkthroughs, instant answers, and seamless escalation pathways.
Adaptive Stage: Instant Answers
Entry Point
User can invoke the Assistant from any page using the nav bar.
Initialization
The Assistant screen loads, populating data that might be relevant.
(common customer queries)
Instant, Inline Answers
The Assistant provides easy access to clear and concise information and relevant knowledge articles.
It repurposes the content relevant to the new page user navigates to.
Follow-up or New Question
User can select a follow-up question directly from the prompt text box or switch to a completely different topic.
Curated Answers
The Assistant streams the answer with clear details and provides a link(s) to the knowledge article.
Future Capabilities
Support Case Creation
The Assistant can identify the issue and it falls under a pre-defined set of rules, it can automatically create a case.
This case would include relevant details gathered during the chat conversation.
Automated Workflows
In scenarios where the action can be automated entirely, Assistant is used to simplify and automate the workflow as much as possible.
Guided Walkthrough
User asks about how to do an action and gets a guided walkthrough.
Escalation
If the issue is complex or requires human interaction, the Assistant can seamlessly transfer the chat to a live customer support agent.
Live Agent Hand-off
User will be routed to Live Chat, with the issue and contextual details carried to the new screen.
Edge Cases
Error handling & Recovery
Assistant recognizes when it's unable to understand the user and take appropriate action, and offers alternative solutions such as browsing knowledge base or access live support.
Server Error
When encountering technical difficulties, the Assistant offers a retry option that allows user to resend their message or question.
