Automate Pipeline SLA Alerts: Connect Your LOS to AI-Drafted Notifications

What This Builds

An automated system that watches your pipeline for files approaching SLA deadlines and sends AI-drafted alert messages to processors and loan officers, without you manually tracking or writing each one. Instead of spending 15-30 minutes each morning reviewing your pipeline and composing individual follow-up emails, the system identifies at-risk files and drafts targeted status request messages automatically, which you review and approve before sending.

The result: your SLA compliance improves, processors get consistent nudges before conditions go stale, and you reclaim nearly an hour each day you were spending on pipeline chasing.

Prerequisites

• Zapier account. $29.99/month (free tier works for basic setup; paid required for multi-step Zaps and higher usage)
• ChatGPT Plus account. $20/month/month (for the AI drafting step)
• Access to your pipeline data in a format Zapier can read (Google Sheets export, Airtable, or direct LOS integration if available)
• Gmail or Outlook connected to Zapier
• Cost: $29.99/month/month + $20/month/month

The Concept

Zapier is a connector: it watches one system for a trigger ("a file has been open for X days without a condition update") and then takes an action in another system ("send a message"). By connecting Zapier to ChatGPT, you can make the action intelligent: instead of sending a generic "please update this file" template, ChatGPT drafts a specific message based on the file's details (loan type, outstanding condition category, how many days past SLA) and addresses it to the right person.

Think of it like a dispatcher who monitors your pipeline queue, spots when something is aging out, and drafts the right message for you to review and send. You stay in control; you just don't have to find the problem or write the message from scratch.

Build It Step by Step

Part 1: Set up your pipeline data source

Zapier needs to read your pipeline data somewhere. Most LOS systems (Encompass, Byte) don't have native Zapier integrations, but you can work around this:

Option A: Google Sheets export (easiest)

1. Export your pipeline to a spreadsheet each morning (most LOS systems support this)
2. Set up a Google Sheet with columns:

• Loan Number
• Borrower First Name (no last name, for internal tracking only)
• Loan Officer Name
• Processor Name
• Processor Email
• LO Email
• Current Status (in underwriting / suspended / approved with conditions)
• Last Condition Update Date
• Days Since Last Update
• Outstanding Condition Category (Income, Credit, Property, Assets, Other)
• Application Date
• Target Close Date

1. Name the sheet: Pipeline Tracker
2. Add a formula column: =TODAY()-[Last Condition Update Date] → this gives you days since last update

Option B: Airtable (more powerful) If your team already uses Airtable for pipeline tracking, Zapier connects to it natively. The setup is the same but you get more filtering options.

Option C: LOS direct integration Some LOS platforms (Encompass in particular) have API access or Zapier integrations via third-party connectors. Check your LOS documentation or IT team, if available, this is the cleanest option.

Part 2: Create your first Zap (the SLA alert)

1. Go to zapier.com and sign in
2. Click + Create → Zaps
3. Name your Zap: "Pipeline SLA Alert. Condition Stale 3+ Days"

Step 1. Trigger: New or Updated Row in Google Sheets

• App: Google Sheets
• Event: New or Updated Spreadsheet Row
• Connect your Google account
• Select your spreadsheet and the Pipeline Tracker sheet
• Click Continue → Test trigger to pull sample data

Step 2. Filter: Only continue if stale

• App: Filter by Zapier
• Condition: Days Since Last Update → Greater Than → 3
• Add another condition: Current Status → Does not contain → Approved
• This ensures you only alert on active files that have stalled, not on files already cleared

Step 3. Action: Draft message with ChatGPT

• App: ChatGPT
• Event: Send Message
• Connect your OpenAI account (the ChatGPT integration uses your OpenAI API key, get this from platform.openai.com → API Keys)
• User message field: Enter this template (Zapier will substitute the actual values from your sheet):

Write a professional email from a mortgage underwriter to a processor requesting a status update on a suspended loan file. Use this information:

Loan number: [Loan Number from sheet]
Processor name: [Processor Name]
Outstanding condition category: [Outstanding Condition Category]
Days since last update: [Days Since Last Update]
Target close date: [Target Close Date]

Requirements:
- Keep it under 100 words
- Professional but direct, not aggressive
- Ask specifically about the [Outstanding Condition Category] conditions
- Include the loan number in the subject line
- Don't include borrower last name or SSN, just the loan number
- End with a clear ask: what do you need and by when

• Model: GPT-4o (more reliable for business writing)
• Click Continue → Test action. Review the draft message

Step 4. Action: Create a draft email in Gmail/Outlook

• App: Gmail (or Outlook)
• Event: Create Draft (NOT Send: you review and send manually)
• To: [Processor Email from sheet]
• Subject: [Loan Number]. Condition Update Needed
• Body: [ChatGPT response from Step 3]
• Click Continue → Test action

1. Click Publish Zap → set it to run Daily (check the schedule trigger option) or leave it event-triggered on sheet updates

Part 3: Set up the daily morning review workflow

With the Zap running, here's your new morning workflow:

1. Open your Google Sheet → update the "Days Since Last Update" column (or have a formula do this automatically)
2. The Zap fires on any row where days > 3
3. Open your Gmail Drafts folder
4. Review each AI-drafted message. They'll be in your drafts, not sent
5. Edit any that need adjustment (usually 1-2 minutes of light editing)
6. Send the ones that are accurate

Time spent: 10-15 minutes reviewing and sending drafts vs. 30-45 minutes identifying, composing, and sending from scratch.

Part 4: Add a second Zap for LO notifications

Repeat the same process for a second Zap focused on LO communication:

Trigger: Same sheet, same stale condition filter ChatGPT prompt variation:

Write a brief email from a mortgage underwriter to a loan officer updating them on a file status. Loan number: [Loan Number]. The file has been suspended for [Days Since Last Update] days pending [Outstanding Condition Category] conditions. Keep it professional, under 75 words, and suggest the LO contact the borrower to help gather [Outstanding Condition Category] documentation.

Action: Draft in Gmail, addressed to [LO Email]

Real Example: Managing a 40-File Pipeline

Setup: Your pipeline has 40 active files. You export to Google Sheets each morning at 8am. The Zap runs at 8:30am against the updated data.

What happens automatically:

• The filter identifies 6 files with conditions stale 3+ days
• For each, ChatGPT drafts a processor email with the specific condition category and loan number
• All 6 drafts land in your Gmail Drafts folder

What you do:

• Open Gmail at 9am. 6 drafts are waiting
• Review each: 4 are good to send as-is, 2 need a small edit ("the income conditions are W-2 verification, not self-employed. Adjust the ask")
• You edit the 2, send all 6 in about 12 minutes

Without this system: You'd open Encompass, sort by last update date, manually identify the 6 files, open each one, check conditions, go back to email, write 6 individual messages, each 5-8 minutes.

Time saved: 12 minutes vs. approximately 50 minutes. Over a week of 5 days: about 3 hours recovered.

What to Do When It Breaks

• Zap fires but ChatGPT draft is off-topic or generic → Check that your Google Sheet fields are mapping correctly into the ChatGPT prompt. Open the Zap editor → Step 3 → look at "Test action" to see what data was actually passed. If the loan number shows as blank, the column mapping is wrong.

• Too many drafts, every file triggers the Zap → Check your filter step. Make sure the "Days Since Last Update" column is formatted as a number, not text. A formula like =TODAY()-C2 returns a number; manually typed "5 days" does not.

• ChatGPT action fails with an API error → Your OpenAI API key may have run out of credits (API usage is billed separately from ChatGPT Plus). Check platform.openai.com → Usage to see your balance. Add credits if needed.

• Processor email addresses in the sheet are wrong or outdated → The Zap will draft to whatever email is in the sheet. This is an input quality problem, make sure your pipeline sheet is maintained. Consider adding a VLOOKUP to pull emails from a separate team directory.

• Google Sheets trigger misses rows or fires duplicate alerts → Switch from "New or Updated Row" to a Scheduled Zap that runs at a fixed time (8:30am) and processes all rows meeting your filter. This is more reliable for daily batches.

Variations

• Simpler version: Skip Zapier entirely. Each morning, paste your pipeline data into ChatGPT and ask: "Here is my pipeline. Write a follow-up email for each file where the last update column shows more than 3 days." Copy each draft into email manually. No automation setup required, takes 10 minutes instead of 2, but no Zapier account needed.
• Extended version: Add a third Zap that pulls a weekly summary from your Google Sheet and sends you a pipeline health report every Friday afternoon: total files, average days to condition clearance, SLA performance by processor. ChatGPT can draft this as a formatted report for your manager.

What to Do Next

• This week: Start with Option A (Google Sheets). Build the first Zap for processor alerts only. Run it for 3 days and review the output quality before expanding.
• This month: Add the LO notification Zap. Track SLA metrics before and after, if condition clearance times improve, you have a concrete metric to show your manager.
• Advanced: If your LOS has an API (Encompass has an API available via Ellie Mae), work with your IT team to replace the Google Sheets data source with a live LOS pull, eliminating the daily export step and making alerts truly real-time.

Advanced guide for mortgage underwriter professionals. This automation requires Zapier and OpenAI API access (billed separately from ChatGPT Plus). Review AI-drafted messages before sending, do not auto-send without human review. Confirm that connecting pipeline data to third-party services (Zapier, OpenAI) complies with your organization's data security and privacy policies.