P
AI for Insurance Claims Adjuster

Claim Intake Automation: Extract Key Facts from Email Notifications

For Insurance Claims Adjusters

Tools: Zapier + ChatGPT API | Time to build: 1-2 hours | Difficulty: Intermediate-Advanced Prerequisites: Comfortable using ChatGPT for basic tasks (Level 3) — see Level 3 guide: "Long Claim File Analysis with Claude Pro"

What This Builds

When a new claim assignment email arrives in your inbox, this automation immediately extracts the key facts (claim number, date of loss, loss type, insured name, location), generates a structured intake note, and optionally drafts your first-contact checklist — all before you've even opened the email. You start every new claim already oriented.

Prerequisites

  • A Zapier account (zapier.com) — free tier works for up to 5 zaps/100 tasks per month; Starter plan at $19.99/month for more
  • An OpenAI account with API access — platform.openai.com (separate from ChatGPT Plus; ~$5-15/month based on usage)
  • Gmail or Outlook email for receiving claim assignments
  • Your new claim assignment emails must come from a consistent email address or follow a consistent format
  • Time needed: 1-2 hours to build; zero ongoing effort after
  • Cost: Zapier ($0-20/month) + OpenAI API (~$5-10/month for typical adjuster usage)

The Concept

Zapier is a connector between apps — it watches for things to happen in one app (like a new email arriving) and then does things in other apps automatically. Think of it as a digital relay race: your email is the starting pistol, ChatGPT runs the analysis lap, and your notes document receives the finished baton.

You set this up once. After that, it runs automatically every time a matching email arrives, even when you're driving to a field inspection or taking a recorded statement.

Build It Step by Step

Part 1: Set Up Your OpenAI API Access

You need API access to ChatGPT — this is different from your ChatGPT Plus subscription.

  1. Go to platform.openai.com and log in (create an account if needed)
  2. Click your profile → API keys → Create new secret key
  3. Name it "Zapier Claim Intake"
  4. Copy the key and save it somewhere safe — you won't see it again. If you lose it, create a new one.
  5. Add a payment method: Settings → Billing → Add payment method. Load $10 to start (this covers approximately 1,000 claim intake analyses)

What you should see: An API key starting with "sk-proj-..." saved to your clipboard.

Part 2: Create Your Zapier Account and Connect Gmail/Outlook

  1. Go to zapier.com and create a free account
  2. Click Create Zap (the automation you're building)
  3. In the trigger section, search for and select Gmail (or Outlook if your carrier uses Microsoft 365)
  4. Select trigger event: New Email Matching Search
  5. Connect your email account by clicking Sign in and allowing Zapier to access your email
  6. In the search field, enter the search terms that identify a new claim assignment email. Examples:
    • from:claims@[yourcarrier].com subject:new assignment
    • subject:claim assigned to you
    • The specific email address or subject line format your carrier uses

What you should see: Zapier finds a recent matching email as a test. If it finds one, you're connected correctly.

Part 3: Add the ChatGPT Analysis Step

After the trigger, click the + button to add an action.

  1. Search for and select OpenAI (ChatGPT) from the app list
  2. Select action event: Send Prompt
  3. Connect your OpenAI account: enter the API key from Part 1
  4. In the Model field, select gpt-4o (most accurate) or gpt-4o-mini (faster and cheaper)
  5. In the User Message field, type:
Copy and paste this
Extract key facts from this insurance claim assignment email and format them as a structured intake note. Email content: [click + and select "Body Plain" from the trigger email step]

Format your response exactly as:
CLAIM INTAKE NOTE
Claim Number: [extract or "Not found"]
Insured Name: [extract or "Not found"]
Date of Loss: [extract or "Not found"]
Type of Loss: [extract or "Not found"]
Location: [extract or "Not found"]
Assigned Adjuster: [extract or "Not found"]
Coverage Line: [extract or "Not found - e.g., auto/property/liability"]
Initial Priority: [assess as High/Medium/Low based on details - High for injuries, large losses, or CAT events]
First Contact Due: [calculate as 24 hours from assignment if High, 48 hours if Medium, 72 hours if Low]

FIRST STEPS CHECKLIST
[ ] Review policy for coverage and limits
[ ] Contact insured within deadline above
[ ] [Add any claim-type-specific first steps based on the type of loss identified]

NOTES
[Any other notable information from the email]

What you should see: A preview showing the formatted intake note generated from your test email.

Troubleshooting: If the email body isn't being mapped correctly, go back to the trigger step and look for "Body Plain" or "Body HTML" as the data fields. Select the one that has actual text content.

Part 4: Save the Output Where You'll See It

After ChatGPT analyzes the email, you need the output to go somewhere useful. Add another action step (+):

Option A: Email it back to yourself

  • App: Gmail (or Outlook)
  • Action: Send Email
  • To: your own email address
  • Subject: "Claim Intake Note — [map Claim Number from your ChatGPT output if possible, or use the trigger email subject]"
  • Body: [map the ChatGPT response text]

Option B: Create a note in Microsoft OneNote or Notion

  • Both have Zapier integrations
  • Creates a new note for each new claim automatically
  • Good if you use a note-taking app for claim management

Option C: Add a row to a Google Sheet

  • Your personal claim tracker
  • Adds each new assignment as a row automatically with key facts extracted

Part 5: Test Your Automation End-to-End

  1. Click Test Zap to run the full automation on your test email
  2. Check your chosen output location — you should see the structured intake note
  3. Review: Did ChatGPT extract the right facts? Is the format useful?
  4. Make adjustments to the prompt if needed
  5. Turn the Zap ON

What you should see: The next time you receive a new claim assignment email, the intake note appears automatically in your inbox (or OneNote/Sheet) within 1-2 minutes.

Real Example: Your First Week of Use

The trigger: Your claim management system sends you an email: "New Claim Assigned: Claim #P2026-8847 — Property Fire Loss — 123 Oak Street, Dallas TX — Date of Loss 03/18/2026 — Insured: Rodriguez, Maria"

What happens automatically (while you're on your way to another inspection): Zapier detects the email → sends it to ChatGPT → ChatGPT produces:

Copy and paste this
CLAIM INTAKE NOTE
Claim Number: P2026-8847
Insured Name: Rodriguez, Maria
Date of Loss: 03/18/2026
Type of Loss: Residential Fire
Location: 123 Oak Street, Dallas, TX
Coverage Line: Property — Homeowners
Initial Priority: High (fire loss)
First Contact Due: 03/19/2026 by [assignment time + 24 hours]

FIRST STEPS CHECKLIST
[ ] Review policy for dwelling coverage, personal property, and additional living expenses
[ ] Contact insured within 24 hours — fire losses require prompt response
[ ] Determine if insured is displaced and needs ALE arrangements
[ ] Assign fire investigation vendor if origin/cause is unclear
[ ] Preserve subrogation rights if arson or third-party cause suspected
[ ] Schedule field inspection within 48 hours

NOTES
Loss appears to be residential structure fire. No injury information mentioned. Verify if fire department report is available.

You arrive at your next inspection already knowing what you're walking into with the Rodriguez claim, what's due when, and what steps to take — without reading the email once.

What to Do When It Breaks

  • Zap doesn't trigger → Check that the search terms in your Gmail trigger actually match the incoming email. Test by searching in Gmail itself first.
  • ChatGPT extracts wrong facts → Your assignment emails may have inconsistent formats. Improve the prompt by adding: "If you cannot find a specific field, write 'Check email manually' rather than guessing."
  • Output is too long or too short → Adjust the prompt. Add "Keep this concise — under 300 words" or "Expand the checklist to include 8-10 steps."
  • API costs more than expected → Switch from gpt-4o to gpt-4o-mini in your Zapier action. It's 15x cheaper with only slightly less accuracy for structured extraction tasks.

Variations

  • Simpler version: Skip Zapier entirely. Just paste your new assignment emails into Claude or ChatGPT manually each morning and run the same extraction prompt. No automation, but the intake note output is identical.
  • Extended version: Add a second Zapier step that creates a calendar event with the first-contact deadline automatically. Requires connecting your Google Calendar or Outlook Calendar to Zapier.

What to Do Next

  • This week: Run the simple version manually first (paste emails into ChatGPT) to verify the prompt extracts the right facts from your carrier's email format.
  • This month: Build the Zapier automation and let it run for 2-3 weeks. Check if the intake notes are accurate and useful.
  • Advanced: Add a second automation that creates a task in your task manager (Todoist, TickTick, Microsoft To Do) with the first-contact deadline automatically set.

Advanced guide for insurance claims adjuster professionals. These techniques use the OpenAI API (requires a separate account from ChatGPT Plus) and Zapier (requires account). API costs are typically $5-15/month for moderate usage. Features and pricing change — verify at platform.openai.com and zapier.com.