Sales ops · Production

ServiceNow builds multi-agent customer success system with LangGraph and LangSmith

The problem

ServiceNow had agents deployed across multiple platform areas with no unified orchestration layer or single source of truth, making it difficult to coordinate complex workflows spanning the entire customer lifecycle.

Workflow diagram · grounded in source
1
Customer signal triggers agents
trigger
“Different triggers activate the appropriate agents based on customer signals and lifecycle stage”
2
Supervisor agent routes to subagents
routing
“The architecture uses a supervisor agent for orchestration, with multiple specialized subagents handling specific tasks”
3
Specialized agents recommend actions
ai_action
“specialized agents determine what actions an Account Executive (AE), seller, or Customer Success Manager (CSM) should take to meet customer requirements”
4
Personalized email and meeting output
output
“automatically drafts personalized emails with relevant information, and schedules meetings between the CSM and customer”
5
LangSmith step-by-step tracing
validation
“LangSmith offers detailed tracing capabilities by providing the input, output, context used, latency, token counts at every step of agent orchestration and helps users to improve the agents performance”
6
LLM-as-judge evaluation
validation
“they define custom scorers based on each agent's specific task. Furthermore, they leverage LLM-as-a-judge evaluators to judge the agent responses”
7
Human feedback integration
human_review
“Leveraging LangSmith's flexibility to collect human feedback and compare prompt versions”
8
Golden dataset feedback loop
feedback_loop
“When prompts meet score thresholds for specific agentic tasks, they're automatically added to the golden dataset”
Reported outcome

ServiceNow built a multi-agent system using LangGraph for orchestration and LangSmith for tracing, which dramatically reduced development friction; the system is currently in the testing phase with QA engineers evaluating agent performance.

Reported metrics
Development frictiondramatically reduced development friction
Reported stack
LangSmithLangGraphLangChainModel Context Protocol (MCP)
Source
https://blog.langchain.com/customers-servicenow/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

ServiceNow built a multi-agent system using LangGraph for orchestration and LangSmith for tracing, which dramatically reduced development friction; the system is currently in the testing phase with QA engineers evalua…

What tools did this team use?

LangSmith, LangGraph, LangChain, Model Context Protocol (MCP).

What results were reported?

Development friction: dramatically reduced development friction (source-reported, not independently verified).

How is this sales ops AI workflow structured?

Customer signal triggers agents → Supervisor agent routes to subagents → Specialized agents recommend actions → Personalized email and meeting output → LangSmith step-by-step tracing → LLM-as-judge evaluation → Human feedback integration → Golden dataset feedback loop.

Related sales ops cases