Intercom Fin AI agent resolving support conversations

L3 · Conditional Verified · Production SaaS Customer Service

Fin (Claude-powered) answers and resolves support queries for 25,000+ Intercom customers; named examples Synthesia, Fundrise, Lightspeed (reported by Anthropic/Intercom)

Problem

Expensive human support and slow responses.

Workflow

Ingest knowledge base
Answer with policy-aware responses
Take actions (refunds, account changes)
Hand off to humans

Human-in-the-loop: Complex queries go to human agents.

Tools & stack

Tools

Intercom Fin
Claude

Integrations

Intercom helpdesk

Frameworks

None listed

Prerequisites

Clean source data and system access for the agent

Risks & limitations

Risks

Vendor-reported metrics, not independently audited

Limitations

Model-vendor case study

Outcomes & evidence

Reported outcomes

51% average resolution rate; up to 86%
Reported by Anthropic/Intercom · Source (anthropic.com) ↗
Lightspeed: up to 65% resolution, agents close 31% more conversations/day
Reported by Anthropic/Intercom · Source (anthropic.com) ↗
Fundrise: >50% of support volume automated in 3 months
Reported by Anthropic/Intercom · Source (anthropic.com) ↗

Evidence

vendor-case-study (named customer) (2026-10-13)
Achieves a 51% average resolution rate out of the box

Vendors & implementers

Intercom

Review history

Last reviewed 2026-10-13 by Research Chief (agent).

2026-10-13: Added in batch-01; source fetched and key figures confirmed.
2026-10-13: Figures attributed to their source; autonomy reviewed.
2026-10-13: Attributed plain-number claims to their source.