ClassPass Hands Most Customer Inquiries to AI With a 24/7 Chat-First Support Model
ClassPass rebuilt its customer support around AI and live chat. The result: 24/7 availability, higher throughput, and costs that don't scale linearly with demand.
The company didn't tack on a bot. It redesigned how work gets done - automated resolution for common issues, real-time AI assistance for agents, and new roles focused on quality and specialization.
From Email Backlog to Chat-First Ops
ClassPass used to rely on email. Agents bounced between knowledge bases, workflow tools, and messaging platforms, which slowed everything down as volume grew.
"We had a growing consumer base and increasing CX costs," said Sarah Vandenbrook, senior manager of CX operations and AI at ClassPass. The team moved to live chat and introduced AI-powered agent assistance to centralize documentation, workflows, and response logic in one place.
"The most noticeable way this changed how our team works is that we are chat-first," Vandenbrook said. Most email agents shifted into live chat, enabling faster response times and true 24/7 coverage.
Automation + Agent Assist in One Flow
Real-time assistance boosted agent confidence and speed, especially while handling multiple chats at once. Automated responses picked up routine questions, cutting queue times and lowering the need for human escalation.
Control lived with CX. "We wanted a tool where my CX team could manage any changes we wanted to make without needing to wait on someone else's timeline," Vandenbrook said. The team could update bot answers and custom workflows immediately.
Redesigning Roles, Not Just Tools
As automation took on repetitive tasks, ClassPass restructured the team. New quality assurance roles reviewed conversations and flagged where automations and workflows needed upgrades.
"We've been able to create new roles," Vandenbrook said. Specialist roles focused on high-friction categories, while the broader team handled general inquiries. Experienced agents concentrated on complex issues instead of simple triage.
Reported Outcomes
- 24/7 chat support coverage.
- Significant increase in automated resolution.
- Decagon cites a 95% reduction in cost per support conversation (partner-reported) and roughly 10x higher automated ticket deflection.
- According to The Information, the support bot now handles a substantial share of inbound inquiries and performs on par with human agents in many categories.
ClassPass shared its approach in a session with Decagon AI, highlighting the importance of unified tooling and CX-owned workflows.
What Support Leaders Can Copy This Quarter
- Go chat-first for high-volume lanes. Keep email for edge cases and regulated flows.
- Centralize knowledge, workflows, macros, and policy logic into a single agent interface.
- Deploy agent assist before full automation. Let AI draft, suggest next steps, and surface policy - agents approve until confidence is proven.
- Define automation boundaries. Start with high-intent, low-risk intents (refund status, booking changes, account updates).
- Create a QA function. Review transcripts daily, tag failure reasons, and feed improvements into bot content and workflows.
- Spin up specialist pods for complex categories (billing edge cases, partner disputes, fraud). Route intelligently.
- Measure what matters: CSAT, first contact resolution, automation containment, median handle time, cost per conversation.
- Keep ownership with CX. Give your team the power to ship content and workflow changes without waiting on engineering.
- Train for concurrency. Coach agents to handle multiple chats with AI guidance, not just faster typing.
Guardrails to Keep Quality High
- Clear fail-safes: confidence thresholds, quick human escalation, and visible handoffs.
- Policy alignment: automate only where policy is explicit and reversible.
- Customer effort: track repeat contact rate and recontact within 72 hours, not just CSAT.
- Ship in slices: roll out by intent, channel, and region; A/B automated vs. assisted paths.
- Content hygiene: weekly updates to macros, bot answers, and decision trees based on QA findings.
- Privacy and security reviews for data access, retention, and PII masking.
Bottom Line
ClassPass didn't chase a shiny tool. It rebuilt support around speed, clarity, and control - then let automation take the repetitive load while experts solved the hard stuff.
If you're reworking your CX stack for the same outcome, a practical next step is to align your team on workflows, QA, and ownership. For structured training on AI automation in support, explore the AI Learning Path for Administrative Assistants, Automation courses, and the AI Automation Certification.
Your membership also unlocks:
