Peter Bar

Peter Bar is a product leader at Intercom behind Fin Voice, an enterprise AI agent that answers customer-support calls, resolves questions, and transfers complex cases to human agents. His work tackles the operational challenges that determine whether automated phone support succeeds: conversational timing, integration with support teams, reliable evaluation, and customer trust.

Bar studied machine learning and computer vision at Imperial College London, completing a master’s thesis on face tracking, and worked on products at Shazam before joining Intercom. As a principal product manager, he helped extend Intercom’s text-based customer-service technology into live voice conversations.

His team built the first version of Fin Voice in approximately 100 days. They initially focused on answering questions from companies’ existing knowledge bases and positioned after-hours support as a low-risk alternative to voicemail. Testing tools, deployment controls, call transcripts, recordings, and escalation settings gave support managers visibility before expanding automation into established workflows.

Defining product ideas

• Voice requires its own conversation design. Silence that users tolerate in chat can feel like a malfunction on the phone. Bar’s team filled retrieval delays with conversational cues, divided complicated answers into spoken steps, and designed around interruptions, pacing, and callers accustomed to rigid phone menus.

• Human handoff determines enterprise adoption. Escalation rules, intelligent routing, identity verification, and conversation summaries help human agents continue a call without asking customers to repeat themselves. Bar treats transparent AI disclosure and operational control as requirements for trust.

• Resolution is the metric that matters. His evaluation framework combines reviewed conversations, logs, recordings, call outcomes, and experimental model-assisted assessment. He favors outcome-based pricing because charging for resolved problems aligns provider incentives with customer results, even though unsuccessful calls leave providers carrying the cost.

• Specialized models make voice agents more capable. Bar helped launch Fin Voice 2, powered by Fin Apex Flash, a customer-service model designed for fast spoken interaction and complex support requests. Intercom’s June 2026 product announcement reported a 24.5% improvement in resolution and responses arriving 0.43 seconds faster, alongside support for refunds, appointments, identity checks, and connected business systems.