Data entry ops · Production

Paradigm builds AI-powered intelligent spreadsheet with LangChain and LangSmith for agent observability and usage-based pricing

The problem

Paradigm needed a sophisticated system to monitor and optimize the performance of hundreds or thousands of AI agents running per-cell spreadsheet tasks, and to implement a precise usage-based pricing model based on actual compute consumption.

Workflow diagram · grounded in source
1
User triggers spreadsheet job
trigger
“users triggering hundreds or thousands of individual agents to perform tasks on a per-cell basis”
2
Schema agent defines columns
ai_action
“Takes in a prompt as context and outputs a set of columns and column prompts that instruct our spreadsheet agents how to gather this data”
3
Plan agent stages and parallelizes tasks
ai_action
“Organizes the agent's tasks into stages given the context of each row of the spreadsheet. This helps parallelize research tasks and reduce latency without sacrificing accuracy.”
4
Swarm agents gather and structure data
ai_action
“orchestrates a swarm of AI agents to gather data, structure it, and execute tasks with human-level precision”
5
LangSmith monitors agent execution
validation
“Track the execution flow of agents, including token usage and success rates”
6
Team refines system via traces
feedback_loop
“the Paradigm team could change the structure of the dependency system, re-run the same spreadsheet job, and assess which system led to the most clear and concise agent traces using LangSmith”
7
Usage-based pricing calibrated
output
“LangSmith gave the Paradigm team perfect context on their agent operations, including the specific tools leveraged, the order of their execution, and the token usage at each step. This allowed them to accurately calculate the cost of dif…”
Reported outcome

LangSmith gave Paradigm full context on agent execution, enabling product and pricing optimization, data quality improvements, and compute cost control.

Reported metrics
Compute costskeeping compute costs low
Latencyreduce latency without sacrificing accuracy
Data qualityimproving data quality
Reported stack
LangChainLangSmith
Source
https://blog.langchain.dev/customers-paradigm/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

LangSmith gave Paradigm full context on agent execution, enabling product and pricing optimization, data quality improvements, and compute cost control.

What tools did this team use?

LangChain, LangSmith.

What results were reported?

Compute costs: keeping compute costs low; Latency: reduce latency without sacrificing accuracy; Data quality: improving data quality (source-reported, not independently verified).

How is this data entry ops AI workflow structured?

User triggers spreadsheet job → Schema agent defines columns → Plan agent stages and parallelizes tasks → Swarm agents gather and structure data → LangSmith monitors agent execution → Team refines system via traces → Usage-based pricing calibrated.

Related data entry ops cases