The scaling customer support problem
Scaling effective customer support was a challenge for eSnipe.com - a fast-growing eBay auction bidding tool.
The company's more than 2,000,000 customers and 100s of product features meant a substantial number of user questions, resulting in a high ticket volume (average: 9,000 tickets/month). Users struggled to self-serve relevant articles in their existing help center, leading to many manual, expensive support tickets and frustrated users.
Led by Winston, a team of developers and customer service managers set out to automate this complex customer workflow.
An AI Agent with a trained knowledge base
With customer experience in mind, Winston and his team used Voiceflow to automate their help center for better customer support. The team designed and deployed a knowledge base-powered agent that integrated with their existing help center interface.
Here's how they did it:
- Curated a Knowledge Base (KB): Winston's team loaded their entire website into a Voiceflow knowledge base, creating a central repository of eSnipe information. The knowledge base consisted of help article URLs, PDF documents, and .txt files.
- Deployed the Knowledge Base Query API: They used the Knowledge Base Query API to enable external access to the curated KB information. This would allow for quick responses to user queries whenever the API was called.
- Connected to a custom interface: A script was created for integrating their AI Agent and eSnipe.com using Voiceflow's API step and CloudFlare. This ensured accurate, rapid answers at their existing help center frontend.
"Customer support ticket volume has decreased by 70% since launching an AI agent in our help center."
Impact on customers & support
Winston and team have seen remarkable impact since launching their AI agent.
- Total ticket volume has drastically decreased. The original 9,000 tickets/month is down to 2,700 tickets/month - a 70% decrease. This is reducing the burden on their support team and signaling that customers are more frequently self-servicing their questions.
- Response times are faster. Customer questions are handled more promptly than the manual ticketing support process.
- Live agents are allocating time better. They can fully focus their time on the 30% of remaining support tickets that need valuable live agent skills and time.
- Support is always user-friendly. The AI-powered responses are trained on the company's content and prompted in their branded voice to make sure the help center agent is an extension of their customer support team.
Hear more from our chat with Winston, including a deep dive into his integration build here.
What's next? Winston and his team are expanding their knowledge base and AI agent build into a chat-based interface, offering features such as account lookup, transaction-specific assistance, and page-specific guidance.