Customer support
 |
VOOZH | about |
|
VOOZH | about |
Build AI support agents with a more human touch and transform your customer experience.
Build AI support agents with a more human touch and transform your customer experience.
Claude can now use Skills—folders with instructions, scripts, and resources—to become a specialist at specific tasks when you need it.
Claude can now use Skills—folders with instructions, scripts, and resources—to become a specialist at specific tasks when you need it.
Claude can now use Skills—folders with instructions, scripts, and resources—to become a specialist at specific tasks when you need it.
Lyft, a global mobility platform across six continents and thousands of cities, handles customer support interactions that range from simple fare questions to complex issues. The company's support team is one of the only teams at Lyft that directly interface with customers, a critically important task to the company's mission of serving and connecting riders and drivers.
In 2023, Lyft faced a critical inflection point. Customer contacts to support were at an all-time high, expectations were rising, and the support experience wasn't keeping pace.
"We were in a really tough spot," said Elyse Hovanesian, Product Lead for AI in Support at Lyft. "Our customers were frustrated. We were pouring a lot of time and energy into creating new products, but they weren't moving the needle."
The problems compounded on both sides of the interaction. Riders and drivers waited 30 to 40 minutes to connect with an agent. When they finally got through, agents in some instances were juggling three or four customers simultaneously, buried in volume and unable to give any single person their full attention. The experience also felt robotic—customers navigated rigid workflows and received copy-pasted responses that didn't address their actual situations.
Lyft was beginning to see signs of agent burnout, and were so buried in volume that they couldn’t put a face and name to those interactions. They didn't have time to actually understand the issue because they were trying to respond and close chats quickly. What got lost was the connection they really wanted to build.
Lyft needed to find a path that would let them both grow the business and improve customer experience.
The team ran comprehensive tests across multiple models, evaluating both quantitative performance, and qualitative factors that mattered for customer experience.
"We tested accuracy of the response and the ability of the AI models to have the tone and persona that represented our brand," Hovanesian said. "In both of those cases, we found that Claude models were best for Lyft’s needs."
The personality dimension proved decisive. Lyft needed an AI that felt like Lyft: confident and capable, but warm and genuine. The team was determined to avoid creating what Hovanesian called "a dreadful chatbot that no one wants to interact with."
The team started with drivers, whose needs are complex and varied. Driver onboarding involves region-specific requirements layered on top of platform-wide policies—exactly the kind of nuanced, educational task where Claude excelled. From there, Lyft expanded to rider support, moving beyond simple Q&A to handling charge disputes, ride experiences, and other issues that previously required human investigation.
"Claude's personality is really what stuck out to me," said Hovanesian. "It felt organic. Our customers were conversing more and opening up about the issues they were having, which then enabled us to solve them better."
Claude now powers Lyft's AI assistant, handling customer support interactions for both riders and drivers. The assistant greets customers by name, investigates their specific situation, and resolves issues in seconds rather than minutes. When cases require human judgment—safety concerns, complex disputes, or situations needing empathy—the system routes customers to human agents who are able to understand issues even faster with AI-generated summaries of the conversation.
The results transformed how Lyft thinks about customer support. Resolution times dropped by over 87%. What once required a 5-to-10-minute wait followed by extended back-and-forth now happens in seconds. The company has saved millions of dollars—funds they reinvested in their support agents.
"Through using Claude, we've saved millions, which we have reinvested in upskilling our customer support agents," said Hovanesian. "We've empowered our agents to focus on those more complex issues that really require human care."
That shows up in programs like Lyft Silver, a service for older riders that pairs a simplified app experience with dedicated human support. Before Claude, the resources for such a program would have been consumed by routine tickets. Now, trained agents can provide high-touch onboarding and a direct phone line, the kind of service this customer segment needs but that wouldn't have been possible at the previous volume.
The shift also changed what it means to be a support agent at Lyft. Agents now handle one customer at a time instead of juggling multiple chats. They have room to understand the full context of an issue, think critically about solutions, and anticipate what else a customer might need.
Support agents now say they can see a real career path at Lyft, signaling a meaningful shift away from burnout that often characterizes support teams.
Looking ahead, Lyft sees Claude handling increasingly complex issues to give their customers time back, while human agents will stay focused on issues requiring human empathy and understanding. The goal is a support experience that feels personal whether interacting with AI or a human. That means responding fast when speed matters, and deeply attentive when connection matters.
"Using Claude has completely changed my perspective for what we can create in support at Lyft," said Hovanesian. "There is so much more that we can do to create a better experience for our customers."
