# How Lyft Increased Its Agent Resolution Rate by 16% with LangSmith and LangGraph

## Executive summary

Lyft addressed the challenge of scaling its customer support agent stack by replacing brittle, deterministic agents with a meta-agent architecture built on LangGraph and LangSmith. This new self-serve platform allows non-engineering personnel (PMs and ops) to deploy new agents via simple configuration and prompting, drastically reducing agent build time from six months to one to two weeks. This accelerated iteration cycle resulted in a 16% increase in the customer resolution rate.

## Key takeaways

- Shift to Self-Service Agent Platform: The team created a platform enabling PMs and ops to build and ship agents using domain knowledge and natural language prompting, minimizing the need for code changes (merely a config change).
- Architectural Improvement via Meta-Agent: The system utilizes a meta-agent where all sub-agents are registered dynamically as nodes in the meta-agent, simplifying the composition and deployment of new agents.
- Significant Operational Gains: The agent build time was reduced from six months to one to two weeks, allowing engineers to focus on complex, foundational improvements while increasing the overall resolution rate by 16%.

## Technical details

- Agent Architecture: The previous agents were basic and deterministic, relying on single LLM calls and internal APIs, lacking sophistication for multi-turn conversations.
- Observability and Debugging: LangSmith was implemented to address the 'observability blind spot,' providing visibility into which component (tool call, prompt, or LLM hallucination) caused a response failure, making debugging and evaluation smooth.
- System Components: The solution involves a meta-agent structure where sub-agents are registered as nodes, managed by configurations in a runtime repository and prompts stored in the LangSmith Prompt Hub.

## Practical implications

- Implementing a meta-agent layer can abstract complex agent logic, allowing non-technical domain experts to contribute to the system's functionality.
- Leveraging dedicated observability tools (like LangSmith) is critical for debugging complex, multi-step LLM pipelines and moving beyond 'black box' deployments.
- Focusing build engineering efforts on creating self-service platforms accelerates the iteration loop and increases the overall product coverage of AI features.

## Topics

Large Language Models (LLMs), Agentic Systems, LangGraph, LangSmith, Build Engineering, Machine Learning Operations (MLOps), LangChain Academy

Source: https://www.youtube.com/watch?v=M9BMTC8o9-w