# How did Anthropic use an AI agent for its own customer support?

> Anthropic chose to buy rather than build, deploying Intercom's Fin (powered by Claude) for customer support. Within weeks Fin reached a 50.8% resolution rate, up from 36% at launch, and saved over 1,700 hours of team time in the first month.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Anthropic (AI company)
- **Industry:** Software, Internet & Media — AI research & products
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution rate: 50.8%, up from 36% at launch [1] (self-reported)
- time saved: over 1,700 hours in the first month [1] (self-reported)

## How it works
1. Fin handles the bulk of repetitive queries from free users over email and Messenger.
2. The team adds content snippets to raise resolution (doubling snippets lifted resolution 5.5% in a week, per source).
3. Complex queries go to human agents.

## Risks & compliance
- Vendor relationship: Fin runs on Anthropic's models

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Intercom, "Build vs buy: Why Anthropic chose Fin" (undated; accessed 2026-10-17) <https://www.intercom.com/customers/anthropic>
   > "50.8% resolution rate: More than half of all customer queries are now resolved by AI, up from 36% at launch."

Canonical: https://agentcasefile.com/cases/anthropic-fin-customer-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile