How to build an Underwriting Submission Assistant

Workflow Overview

1. Customer Data Collection

The workflow begins by collecting all necessary information from the customer through a series of input nodes:

• Device Brand (in-0): Customer enters the brand (e.g., Apple, Samsung).

• Device Model (in-1): Customer enters the model (e.g., iPhone 15).

• Serial/IMEI (in-2): Customer enters the device’s serial number or IMEI.

• Purchase Date (in-3): Customer enters the date of purchase.

• Purchase Price (in-4): Customer enters the price paid for the device.

• Customer Name (in-5): Customer enters their full name.

• Customer Email (in-6): Customer enters their email address.

• Customer DOB (in-7): (Optional) Customer enters their date of birth.

• Customer Address (in-8): (Optional) Customer enters their address.

• Proof of Purchase Upload (doc-0): Customer uploads a receipt, invoice, or carrier bill. The file is processed with OCR to extract text.

2. Eligibility & Document Validation

• Node: Eligibility and Document Validation (llm-0)

• What it does:
  This AI node reviews all the collected information and the uploaded proof of purchase. It checks:

  • If the device is supported (smartphone, tablet, laptop, wearable)

  • If the purchase is within 30 days

  • If the receipt is valid and matches the device

  • If the serial/IMEI is valid

  • If the device value is under $2,500

  • If the device is not already damaged

  Output:

  • Eligibility status (Eligible / Ineligible / Needs more information)

  • Short explanation referencing the rules

  • List of missing or unclear items (if any)

3. Risk Tier Assignment

• Node: Risk Check (llm-1)

• What it does:
  Using the validated data, this AI node assigns a risk tier (A, B, C, or D) based on:

  • Device type, age, value, brand/model

  • Document quality

  • Any prior claims

  Output:

  • Assigned risk tier

  • Rationale (2–3 short bullets)

  • Route (Auto-bind or Manual review)

4. Pricing & Terms Calculation

• Node: Pricing and Term Calculation (llm-2)

• What it does:
  Based on the risk tier and device value, this AI node calculates:

  • Monthly premium

  • Deductible

  • Coverage limit

  • Maximum claims per year

  • Waiting period

  • Bind decision (Auto-bind or Manual review)

  • Any special notes

5. Policy Document Generation

• Node: Create Google Doc (action-0)

• What it does:
  This action node creates a new Google Doc containing a summary of the customer’s insurance policy.

  • The document is named using the customer’s name.

  • The content includes a personalized message and the calculated policy details (from the Pricing and Term Calculation node).

  • The document is created in Google Drive, and a link to the file is generated.

6. Customer Notification

• Node: Send Email 1 (action-1)

• What it does:
  This action node sends an email to the customer using the Gmail API.

  • The recipient is the customer’s email address.

  • The subject is set to “Next Step: Activate Your Device Coverage.”

  • The body of the email contains the policy details and coverage summary, as generated by the Pricing and Term Calculation node.

  • Optionally, you could attach the generated Google Doc or other files if needed.

7. Agent/Workflow Output

• Node: Output (out-0)

• What it does:
  This node displays the final summary or status to the agent or user running the workflow, showing the results of the policy calculation and approval process.

How Data Flows

• All customer inputs and uploaded documents are fed into the first AI node for eligibility and document validation.

• The output is passed to the next AI node for risk assessment.

• The risk tier and validated data are then used by the third AI node to calculate pricing and terms.

• The results are used to generate a Google Doc and send an email to the customer.

• The final output is shown to the agent for review or record-keeping.

In summary:
This workflow automates device insurance underwriting by collecting customer/device info, validating eligibility, assigning risk, calculating pricing, generating policy documents, and notifying the customer—all with minimal manual intervention.