Eye On A.I. #295 Fergal Reid: Why Your Bots Fail and How Agents Fix Your Customer Support
15 snips
Oct 19, 2025 
In this engaging conversation, Fergal Reid, Chief AI Officer at Intercom and a PhD in machine learning, discusses the challenges of traditional chatbots and introduces Finn, an advanced AI agent designed for customer support. He explains how Finn handles complex tasks like refunds and returns while seamlessly integrating with existing systems. Fergal also highlights the importance of metrics like resolution rate and offers insights on effectively onboarding AI solutions. Their approach balances configuration with a standardized core, ensuring reliability across multiple communication channels.
AI Snips
Chapters
Transcript
Episode notes
Replace BPOs For Routine Support
- Use agents to replace outsourced tier-one BPO work and redeploy human teams to higher-value tasks.
- Fergal says many customers ended BPO contracts after deploying Fin for frontline queries.
Design Reliable Human Handoffs
- Build robust human escalation and handoff flows rather than assuming full automation.
- Intercom invested years in graceful escalation and recently shipped a model that detects when to escalate.
Resolution Rate Is The Real KPI
- Resolution rate is the north-star metric for agents; Intercom raised Fin's resolution from ~35% to ~65% via continuous A/B testing.
- Fergal credits a dedicated team and scientific experimentation for steadily improving performance.