AI-Powered Customer Intelligence for Insurance Claims

AI-Powered Customer Intelligence for Insurance Claims

This workflow brings together insurance claim data, customer survey responses, and garage repair quotations to generate a comprehensive view of each insurance case. It works by:

• Importing and joining claim records, customer details, and accident images from a Snowflake databases; garage repair quotations from local PDF files; and customer survey responses from Google Sheets.

• Formatting and combining this information into structured messages, including both text and images, to create prompts for a multimodal LLM.

• Sending these prompts to the LLM to extract structured information about the claim and repair costs (e.g., accident gravity, car damages, expected and actual repair costs) and to assess customer sentiment and key aspects from survey answers.

• Calculating differences between insurance coverage and both expected and actual repair costs.

• Aggregating results and presenting them in an interactive dashboard for customer intelligence analysis.

Connect to Hub and access credentials

Connect to Snowflake WH, select insurance claims and customer tables, join them, map data types and sort Customer ID

Read garage quotations as local PDFs

Connect to Google Sheet, read customer survey responses

Authenticate to OpenAI and select LLM

Create Prompt (only text)

Create Prompt (text + images)

Prompt LLM and get responses as structured output

Postprocess responses, compute expected vs. actual costs and join data

Customer Intelligence View

Replicate execution

If you don't have access to the

Google Sheet, read customer

survey responses using the

Table Reader node

Replicate execution

If you don't have access to

Snowflake data, read insurance claims

and customer details using the

Table Reader node

Replicate execution

Provide your own OpenAI API key credentials in the Credentials Configuration node and pass it on to the OpenAI Authenticator node