Customer support · Production

Gamma scales to 50M users with 20 agents using Intercom Fin AI

The problem

As Gamma's user base surpassed 50 million, their lean support team faced inbound volumes regularly hitting 30,000 conversations per month, a median first-response time that had climbed to 90 minutes, CSAT that had dropped to around 69%, and no practical path to 24/7 multilingual coverage without dramatically growing headcount.

First attempt

An initial trial of Fin in early 2024 did not deliver the expected results and was deferred; a subsequent attempt to migrate to a different vendor stalled due to roadblocks around support and transparency.

Workflow diagram · grounded in source
1
Inbound conversation arrives
trigger
“Inbound conversation volumes regularly hit 30,000 per month, with requests coming from across the globe in multiple languages and across time zones”
2
Fin engages every conversation
ai_action
“Fin is now involved in 100% of inbound support conversations”
3
Route by tier and sentiment
routing
“Free users are routed directly to Fin, while custom escalation workflows ensure that Fin hands off only when appropriate – routing conversations based on user tier, payment signals, and sentiment”
4
Autonomous end-to-end resolution
ai_action
“resolves 75% of them end-to-end”
5
Real-time translation
ai_action
“It automatically translates conversations in real time, helping customers in their preferred language without the need for multilingual agents”
6
Stripe billing integration
integration
“handle refund requests and payment failures through Stripe integrations”
7
Escalation to human agents
human_review
“Fin escalated a case when it saw a user struggling to pay – even though their metadata marked them as free”
Reported outcome

Fin now handles 100% of inbound conversations, resolves 75% end-to-end, and delivers over 18,000 resolutions each month, reducing manual handling from 94% to just 24% while holding CSAT steady at 84% — all with a team of only 20 outsourced agents supporting 50 million users.

Reported metrics
Fin involvement in conversations100%
Conversation resolution rate75%
Monthly resolutions deliveredover 18,000
Manual handling rate before94%
Show all 13 reported metrics
Fin involvement in conversations100%
conversation resolution rate75%
monthly resolutions deliveredover 18,000
manual handling rate before94%
manual handling rate after24%
CSAT current84%
CSAT before (dropped to)around 69%
median first-response time before90 minutes
monthly inbound conversation volume30,000 per month
initial holiday rollout resolution rate50%
initial holiday rollout CSAT75%
total users supported50 million users
human support agent count20 outsourced agents
Reported stack
IntercomFinSlackStripe
Source
https://www.intercom.com/customers/gamma
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Fin now handles 100% of inbound conversations, resolves 75% end-to-end, and delivers over 18,000 resolutions each month, reducing manual handling from 94% to just 24% while holding CSAT steady at 84% — all with a team…

What tools did this team use?

Intercom, Fin, Slack, Stripe.

What results were reported?

Fin involvement in conversations: 100%; Conversation resolution rate: 75%; Monthly resolutions delivered: over 18,000; Manual handling rate before: 94% (source-reported, not independently verified).

What failed first in this deployment?

An initial trial of Fin in early 2024 did not deliver the expected results and was deferred; a subsequent attempt to migrate to a different vendor stalled due to roadblocks around support and transparency.

How is this customer support AI workflow structured?

Inbound conversation arrives → Fin engages every conversation → Route by tier and sentiment → Autonomous end-to-end resolution → Real-time translation → Stripe billing integration → Escalation to human agents.

Related customer support cases