Customer support · Production

OpenRecovery builds a multi-agent AI recovery assistant with LangGraph and LangSmith

The problem

Addiction recovery support faces a gap between costly inpatient care and generic self-help programs, leaving those struggling with addiction without accessible, expert-level, personalized guidance.

Workflow diagram · grounded in source
1
User sends text or voice message
trigger
“provides personalized, 24/7 support via text and voice”
2
Context switch between agents
routing
“LangGraph also enables smooth context switching between different agents within the same conversation. Users can transition from general chat to specific recovery work without disruption”
3
Specialized recovery node processing
ai_action
“the team specialized nodes in LangGraph, each with tailored prompts for specific stages of the recovery process, such as step work or fear inventory”
4
AI gauges readiness for confirmation
ai_action
“the AI encourages deeper introspection by prompting users, much like a sponsor or therapist would. It gauges when enough information has been collected and requests human confirmation when needed for better accuracy and understanding”
5
Human confirmation and editing
human_review
“users can edit AI-generated summaries or tables, allowing them to verify the accuracy of their personal information and maintain control over their data”
6
Failure identification and few-shot correction
feedback_loop
“they can quickly add new few-shot examples to the dataset in LangSmith, re-index it, and test the same question to verify the improvement. This enforces a cycle of continuous improvement”
Reported outcome

OpenRecovery built a sophisticated, scalable multi-agent mobile application that adapts to individual users' recovery journeys, with human-in-the-loop features for trust and LangSmith-accelerated development.

Reported metrics
Development process speedaccelerated OpenRecovery's development process
Testing robustnessadded robustness to their testing
Reported stack
LangGraphLangSmithLangGraph PlatformLangGraph Studio
Source
https://blog.langchain.dev/customers-openrecovery/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

OpenRecovery built a sophisticated, scalable multi-agent mobile application that adapts to individual users' recovery journeys, with human-in-the-loop features for trust and LangSmith-accelerated development.

What tools did this team use?

LangGraph, LangSmith, LangGraph Platform, LangGraph Studio.

What results were reported?

Development process speed: accelerated OpenRecovery's development process; Testing robustness: added robustness to their testing (source-reported, not independently verified).

How is this customer support AI workflow structured?

User sends text or voice message → Context switch between agents → Specialized recovery node processing → AI gauges readiness for confirmation → Human confirmation and editing → Failure identification and few-shot correction.

Related customer support cases