How to build an Underwriting Submission Assistant

TL;DR

• What it does: Captures an underwriting submission, extracts key fields with AI, checks Salesforce for matches, creates a Notion summary, emails stakeholders, and displays results.

• Who it’s for: Underwriting Ops, Brokers, MGA/MGU teams, Carrier intake teams, RevOps.

• Time to value: ~30–60 min setup; 2–4 min per submission after.

• Output: Structured summary (applicant, coverage, risk), Salesforce matches, Notion page, notification email, on-screen status.

• Fit: Works with text + file inputs; integrates with Salesforce, Notion, and Gmail; exports viewable in-app.

Common Pain Points of Underwriting

• Submissions arrive unstructured; key details scattered or missing.

• Manual rekeying into CRM/underwriting systems creates errors and delays.

• Context lives across email threads and docs with poor traceability.

• Slow internal comms; stakeholders don’t get timely, consistent summaries.

• No single source of truth for each submission.

What the Agent Delivers

• AI-extracted, structured summary of applicant, requested coverage, and risk details.

• Salesforce lookup to surface existing accounts/contacts and avoid duplicates.

• Notion page auto-created as a canonical record with the structured summary.

• Email notification with the summary and Notion link to keep stakeholders aligned.

• Unified output view showing summary, CRM matches, Notion URL, and send status.

Step-by-Step Build on StackAI:

1) Text Input (in-0)

• Purpose: Capture underwriting details (applicant info, requested coverage, risk details, etc.).

• How it works: User types or pastes the submission into the input.

2) OpenAI (LLM) Summarizer (llm-0)

• Purpose: Extract and summarize key fields.

• How it works: Converts raw text into a structured summary (e.g., applicant, coverage, risk).

• Prompt: Instructed to output a concise, structured summary.

3) Salesforce Query (action-0)

• Purpose: Check for existing applicant/risk data.

• How it works: Uses the AI summary to query Salesforce for accounts and contacts.

  
  

  • Connection: Login as Domain: ae5945f1-d45c-4498-9f9d-a8f17b385a65

  
  

• Inputs:

  
  

  • Account/Contact schema.

  • Query built from applicant details in the summary.

4) Create Notion Page (action-1)

• Purpose: Create a canonical submission record.

• How it works: Generates a Notion page from the AI summary.

  
  

  • Connection: Notion 2da6300c-456f-46c2-9d6b-85d2f3373d6b

  • Parent Page: Set your Notion parent page ID.

  
  

• Inputs:

  
  

  • Parent page ID (<YOUR_PARENT_PAGE_ID>).

  • Title: “Underwriting Submission Summary”.

  • Content: AI summary.

5) Send Email (action-2)

• Purpose: Notify stakeholders with the summary + Notion link.

• How it works: Sends an email with both the structured summary and Notion URL.

  
  

  • Connection: Gmail e1d23fac-2626-4526-ad61-ecfd36438ea9

  
  

• Inputs:

  
  

  • Recipient: <RECIPIENT_EMAIL>

  • Subject: “Underwriting Submission Summary”

  • Body: Summary + Notion link

6) Output (out-0)

• Purpose: Display everything to the user.

• How it works: Shows the AI summary, Salesforce results, Notion link, and email status.

Data Flow

1. User Input → AI Summarizer: Raw submission becomes a structured summary.

2. AI Summary → Salesforce Query: Find existing accounts/contacts to prevent duplicates.

3. AI Summary → Notion Page: Create a canonical record.

4. AI Summary + Notion Link → Email: Notify stakeholders.

5. All Results → Output: Display everything in one place.

Node Table