Customer support · Production

Numan scales safe, AI-first patient support to 47% resolution rate with Fin AI Agent

The problem

Numan's support volume surged to nearly 70,000 conversations a month, spanning delivery queries to sensitive clinical topics. Their existing platform could not scale intelligently, requiring the team to manually manage every type of query while ensuring medical questions were safely routed to licensed clinicians.

First attempt

Numan's previous customer service platform lacked the intelligence to handle growing volumes, and competing AI tools they evaluated failed in real-world testing despite performing well in demos.

Workflow diagram · grounded in source
1
Customer query received
trigger
“queries ranging from delivery updates to sensitive questions about treatment plans”
2
Fin generates response from knowledge base
ai_action
“Fin was the only one that gave high-quality responses based on our actual help center and blog content – with almost no setup”
3
Clinical queries routed to clinicians
routing
“the team implemented a set of clear rules to instantly hand over any conversation involving medication, treatment decisions, or follow ups regarding side effects to a clinician. These queries would go straight to licensed clinicians or m…”
4
Non-clinical queries resolved by Fin
output
“resolve a large share of conversations instantly”
5
Human review and ongoing audit
human_review
“The team began by manually reviewing 100% of Fin interactions to ensure quality, accuracy, and safety. "We were cautious from day one," says Rhidian Boobier, Head of Customer Operations at Numan. "We wanted to be sure that Fin could oper…”
Reported outcome

Fin autonomously resolved 47% of conversations, saved 19,000 hours annually (the equivalent of nine full-time agents), and maintained a 90% CSAT score on par with human agents.

Reported metrics
Hours saved annually19,000 hours
Full-time agent workload equivalentequivalent to the workload of nine full-time agents
Cost saved per resolved queryaround four pounds
Fin resolution rate47%
Show all 6 reported metrics
hours saved annually19,000 hours
full-time agent workload equivalentequivalent to the workload of nine full-time agents
cost saved per resolved queryaround four pounds
Fin resolution rate47%
CSAT for Fin90%
monthly support conversation volumenearly 70,000 conversations a month
Reported stack
FinFin GuidanceIntercom
Source
https://www.intercom.com/customers/numan
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Fin autonomously resolved 47% of conversations, saved 19,000 hours annually (the equivalent of nine full-time agents), and maintained a 90% CSAT score on par with human agents.

What tools did this team use?

Fin, Fin Guidance, Intercom.

What results were reported?

Hours saved annually: 19,000 hours; Full-time agent workload equivalent: equivalent to the workload of nine full-time agents; Cost saved per resolved query: around four pounds; Fin resolution rate: 47% (source-reported, not independently verified).

What failed first in this deployment?

Numan's previous customer service platform lacked the intelligence to handle growing volumes, and competing AI tools they evaluated failed in real-world testing despite performing well in demos.

How is this customer support AI workflow structured?

Customer query received → Fin generates response from knowledge base → Clinical queries routed to clinicians → Non-clinical queries resolved by Fin → Human review and ongoing audit.

Related customer support cases