Customer support · Production

Rakuten Group builds LLM-powered AI products for business clients and employees using LangChain and LangSmith

The problem

Rakuten Group saw an opportunity to augment client and employee support at scale with AI, as large-company teams were developing ideas independently with no systematic way to identify and share effective approaches across the organization.

Workflow diagram · grounded in source
1
Client seeks business support
trigger
“When clients onboard their businesses, they receive support from a dedicated onboarding consultant, and once live, continue to get help from them.”
2
AI Analyst: market intelligence
ai_action
“Rakuten AI Analyst which acts as a research assistant, providing valuable market intelligence. This helps clients get business insights backed by relevant data and charts.”
3
AI Agent: self-serve support
ai_action
“Rakuten AI Agent which supports clients in getting faster, self-serve customer support for their questions related to listing and transacting on the marketplace.”
4
AI Librarian: documentation Q&A
ai_action
“Rakuten AI Librarian which summarizes all of the client's documentation to answer questions from the client's end customers and prospects in real time.”
5
Employee chatbot building
ai_action
“leveraged LangChain's OpenGPTs package to deliver an employee empowerment experience, in which teams could build their own chatbots over internal documentation to help with knowledge management and employee enablement. It only took three…”
6
LangSmith prompt and eval loop
feedback_loop
“By using LangSmith Hub, we could distribute the best prompts and promote collaboration across teams. By using LangSmith Testing and Eval with our custom evaluation metrics, we can run experiments on multiple approaches (models, cognitive…”
Reported outcome

Rakuten deployed a suite of AI products for business clients and an internal employee chatbot platform built by three engineers in one week; the platform is planned to reach 32k employees with an aim to improve productivity by 20%.

Reported metrics
Time to build initial employee chatbot platformone week
Engineers to build initial platformthree engineers
Planned employee rollout32k employees
Target productivity improvement20%
Reported stack
LangChainLangSmithOpenGPTs
Source
https://blog.langchain.dev/customers-rakuten/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Rakuten deployed a suite of AI products for business clients and an internal employee chatbot platform built by three engineers in one week; the platform is planned to reach 32k employees with an aim to improve produc…

What tools did this team use?

LangChain, LangSmith, OpenGPTs.

What results were reported?

Time to build initial employee chatbot platform: one week; Engineers to build initial platform: three engineers; Planned employee rollout: 32k employees; Target productivity improvement: 20% (source-reported, not independently verified).

How is this customer support AI workflow structured?

Client seeks business support → AI Analyst: market intelligence → AI Agent: self-serve support → AI Librarian: documentation Q&A → Employee chatbot building → LangSmith prompt and eval loop.

Related customer support cases