Sales ops · Production

Unify builds account qualification agents powered by LangGraph and LangSmith

The problem

Go-to-market teams needed a systematic way to research and qualify prospect accounts based on custom criteria, a task requiring web search, website visits, and multi-source synthesis that was not yet automated.

First attempt

The initial agent version produced inconsistent results and made it difficult to analyze reasoning. The early UX showed only a spinner during agent runs, which became painful as runtimes grew.

Workflow diagram · grounded in source

Account qualification request

trigger

“Given a company and a set of questions and criteria, the agent performs some research and decides whether they are "qualified" or not”

AI plan generation

ai_action

“The first step involves using a large model to generate a plan. In our testing, mainline models like gpt-4o did not construct particularly comprehensive plans without very specific prompting. The best results we obtained at this stage we…”

Reflect and web research loop

ai_action

“the agent then begins looping between a "reflect" step and tool calling”

Qualified/not-qualified output

output

“decides whether they are "qualified" or not”

LangSmith experiment evaluation

feedback_loop

“We're able to run a new agent version on hundreds of examples and quickly compare it against previous versions on a given dataset with very little in house ML infra work”

Reported outcome

Unify evolved to a plan-reflect-tools architecture with parallelized tool calls for speed and a real-time step-by-step UI showing the agent's decision-making, with LangSmith enabling experiment comparison across hundreds of examples with minimal in-house ML infrastructure.

Reported metrics

O1-preview planning step latencyup to 30-45 seconds

Agent runtime improvement from parallelized tool callsone of the biggest speed boosts

Evaluation dataset scalehundreds of examples

Reported stack

LangGraphLangSmitho1-previewGPT-4o3.5 Sonnet

Source

https://blog.langchain.dev/unify-launches-agents-for-account-qualification-using-langgraph-and-langsmith/

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Unify evolved to a plan-reflect-tools architecture with parallelized tool calls for speed and a real-time step-by-step UI showing the agent's decision-making, with LangSmith enabling experiment comparison across hundr…

What tools did this team use?

LangGraph, LangSmith, o1-preview, GPT-4o, 3.5 Sonnet.

What results were reported?

O1-preview planning step latency: up to 30-45 seconds; Agent runtime improvement from parallelized tool calls: one of the biggest speed boosts; Evaluation dataset scale: hundreds of examples (source-reported, not independently verified).

What failed first in this deployment?

The initial agent version produced inconsistent results and made it difficult to analyze reasoning.

How is this sales ops AI workflow structured?

Account qualification request → AI plan generation → Reflect and web research loop → Qualified/not-qualified output → LangSmith experiment evaluation.