G2

Vikash K.

SWE

Insurance

Mid-Market (51-1000 emp.)

9/4/2026

"LangChain Streamlined RAG and model swapping for our insurance claim application."

4.5/5

What do you like best about Langchain?

As an AI engineer, I worked on multiple AI projects and initially performed normal LLM calls, so no external framework was required. Once we started building better AI applications, we needed a framework to handle multiple LLM models and required built-in methods & models.

So Langchain was already a popular framework, and we started using it in our Insurance Claim processing service, where adjusters switch between different models easily. While in our claim files, with the help of Lngchain, we routed to pricing agents.

It integrates smoothly with our vector storage in Pinecone.

Langchain-structured output parser with Pydantic has made line item extraction highly reliable. Review collected by and hosted on G2.com.

What do you dislike about Langchain?

We found some methods difficult to debug, like RecursiveCharacterTextSplitter.

For us, while price matching with candidates via multiple loops seems painful without LangSmith tracing.

Most of the time, while development, found methods are frequently updated, so it breaks sometimes in the latest package version. Review collected by and hosted on G2.com.

What problems is Langchain solving and how is that benefiting you?

Currently, in our project, Langchain is orchestrating our entire RAG pipeline.

Major benefit is model flexibility; our admin can easily swap between different LLMs for specific line-item it's boosted our application's performance.

While Normal LLM calls format prompts according to the model was challenging, so we used PromtTemplate and ChatPromptTemplate. Review collected by and hosted on G2.com.