CHALLENGES

Retell AI needed smarter automation to handle growing support volume while keeping engineers focused on product development.

‍

Retell AI is redefining the call center with AI-native phone agents that answer calls in real time for both inbound and outbound use cases. As Retell AI grew to 200,000 users and scaled past 30 million calls each month, the support team relied heavily on human agents and on-call product engineers to handle everything from simple documentation questions to complex technical debugging.

‍

The team was operating over capacity with their existing support approach. Simple L1 questions that could be answered from existing documentation were pulling engineers away from core product work, creating bottlenecks and growing backlogs. As a fast-scaling startup, they needed a support platform that could grow with them by handling the repetitive work and escalating complex issues appropriately.

‍

“We were essentially operating over capacity. Our on‑call engineers were taking all the tickets and it was overwhelming."

Prabir Vora, Founding Chief of Staff

‍

SOLUTIONS

Pylon's AI Agents became Retell's intelligent triage system, automating routine work while escalating complex issues.

‍

Retell turned to Pylon's AI Agents to create an intelligent triage system that could handle routine questions while ensuring complex issues reached the right engineers with full context. The goal wasn't full automation, but strategic automation that maximized engineering efficiency.

‍

The team specifically implemented:

  • AI Agents with Knowledge Base Integration: Pylon ingested Retell's comprehensive documentation to automatically resolve common L1 questions about API usage, implementation, and troubleshooting the repetitive queries that were consuming engineering time.
  • Structured Runbooks for Complex Workflows: The team built AI runbooks for specialized processes like scam detection checks, BAA/DPA form routing, and multi-language support, ensuring consistent handling of structured L2 workflows.
  • Agents for Ticket Pre-work: For voice-related issues, AI agents now collect critical details like call IDs upfront, so engineers can jump straight into debugging instead of going back-and-forth for basic information.
  • Knowledge Gap Analysis: Pylon's knowledge gap feature helped identify recurring topics that weren't covered in documentation, enabling continuous expansion of AI capabilities.

‍

“On‑call has become a lot easier. The AI does the pre‑work for us, so we can jump straight into triage.”

Prabir Vora, Founding Chief of Staff

‍

RESULTS

Retell achieved 80% L1 automation and freed 12-15 engineering hours per week for product development.

‍

By systematically training and expanding their AI agents over several months, Retell transformed their support efficiency. Engineers now spend dramatically less time on routine inquiries and more time on high-impact product work that drives growth. This shift shows up clearly in the numbers:

‍

  • 80% of L1 tickets automated, freeing engineers from repetitive questions.
  • 50% of L2 workflows supported by AI, with humans stepping in for final actions.
  • 12–15 engineer hours saved per week through reduced back‑and‑forth and better upfront context.
  • 20% of email tickets receive an AI response, with ~50% of those fully resolved.

‍

Beyond the metrics, Retell saw comprehensive operational improvements: engineers eliminated context switching through upfront information gathering, faster triage with call IDs and relevant details collected before handoff, consistent routing for specialized workflows like compliance and partnerships, and continuous optimization through knowledge gap identification.

‍

The impact on team morale was equally significant. Engineers report dramatically improved quality of life during on-call rotations, with AI handling the repetitive work that previously made support feel overwhelming. With Pylon's AI agents as their foundation, Retell is well positioned to continue scaling efficiently while maintaining exceptional customer support quality.

‍

"Pylon's AI agents now handle a significant portion of tickets and take 12-15 hours of repetitive work off engineers each week. That time goes back into building product, and our customers get faster, more accurate answers."

Prabir Vora, Founding Chief of Staff