Customer support · Production

AI-first by design: How Anthropic transformed support operations with Fin

The problem

Anthropic's support operation started with a single person managing rapidly growing conversation volume across a diverse user base — from free consumers to enterprise API customers — with no scalable system and constant trade-offs being made.

Workflow diagram · grounded in source
1
Support conversation arrives
trigger
“Its users range from free consumers to enterprise API customers – each with different needs, expectations, and levels of technical complexity”
2
Fin resolves conversation end-to-end
ai_action
“many subscription, billing, and account workflows move through Fin end-to-end – from retrieving live account data to translating backend error codes and escalating only when necessary”
3
Route complex cases to humans
routing
“Clear routing logic so Fin handles volume at the front line, and humans own enterprise and edge cases”
4
Specialists handle escalated issues
human_review
“Specialists now focus on the parts of support that require human judgement, like deep enterprise issues, compliance edge cases, and proactive outreach”
5
Claude-aided gap analysis
feedback_loop
“Using Claude to summarize chats, cluster gaps, and accelerate content creation”
6
Weekly reviews and content sprints
feedback_loop
“Weekly reviews of unanswered questions to identify gaps and inform content updates”
Reported outcome

Fin now resolves 57% of the conversations it touches, generating around 40 to 50 thousand resolutions per month, while the human team has shifted to high-value enterprise and compliance work.

Reported metrics
Fin conversation resolution rate57%
monthly resolutions via Fin40 to 50 thousand resolutions a month
Resolution rate increase during hack week10%
early conversation share handled by Finnearly half of our conversations
Reported stack
FinClaude
Source
https://www.intercom.com/customers/anthropic-transformation
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Fin now resolves 57% of the conversations it touches, generating around 40 to 50 thousand resolutions per month, while the human team has shifted to high-value enterprise and compliance work.

What tools did this team use?

Fin, Claude.

What results were reported?

Fin conversation resolution rate: 57%; monthly resolutions via Fin: 40 to 50 thousand resolutions a month; Resolution rate increase during hack week: 10%; early conversation share handled by Fin: nearly half of our conversations (source-reported, not independently verified).

How is this customer support AI workflow structured?

Support conversation arrives → Fin resolves conversation end-to-end → Route complex cases to humans → Specialists handle escalated issues → Claude-aided gap analysis → Weekly reviews and content sprints.

Related customer support cases