How to Automate Inbound Email Classification with GPT and Slack Alerts Using Make

Introduction

Inbound email triage is a recurring bottleneck for many teams. You get inquiries from prospects, support requests, and partner updates, all with different urgency. Reading every message, deciding what matters, and routing it to the right channel wastes minutes or even hours each day. This automation solves that pain by using OpenAI's GPT to classify each inbound email and Make to orchestrate the flow. High priority items land in Slack for immediate action, while all messages get logged for audit and improvement over time. By the end, you’ll know how to wire Gmail, OpenAI, Slack, and Make into a reliable end-to-end AI-assisted workflow.

Note: no related posts found for this exact integration in our catalog. You can still benefit from this practical blueprint and tailor it to your stack. If you want this done for your team, Olmec Dynamics can design and deploy it for you. See details at the end of this post.

What You'll Need

• A Make.com account with scenario editing privileges (paid plan recommended for AI modules and higher task limits).
• A Gmail account that will receive inbound emails to classify.
• An OpenAI API key (or access via Make's OpenAI integration) to run GPT classification.
• A Slack workspace with a channel for alerts and the Make app installed in Slack.
• Optional: Google Sheets or Notion for a lightweight audit log (we’ll show a log option in the steps).
• A descriptive label or filter in Gmail to identify messages that should be auto-classified (for example label: ai-classify).

How It Works (The Logic)

Trigger: a new email arrives in Gmail that matches your classification label.

Action: Make sends the email subject and body to OpenAI GPT for classification. The model returns a structured JSON with fields like priority, category, and a short summary.

Decision: if priority is High, post a concise alert to a Slack channel with sender, subject, and a short summary; if Medium or Low, log the entry and optionally prepare follow-up tasks without alerting Slack.

Log: write the result to a Google Sheet or your chosen log so you can track accuracy and improve the prompt over time.

This pattern keeps urgent requests visible while preserving a record of everything for reporting and training data. If you want to see similar patterns in action, our team has implemented end-to-end AI classification and routing in other contexts Olmec Dynamics.

Step-by-Step Setup

1. Create a new scenario in Make.com and name it Inbox AI Classification. Connect the four apps: Gmail, OpenAI, Slack, and Google Sheets (optional log).

2. Gmail module: add a watcher or watch emails module set to the inbox with a filter for the label you decided on (for example, label ai-classify). Configure it to trigger when a new message arrives. Map fields you need later: from, subject, date, and snippet/body excerpt.

3. OpenAI module: add an action to call GPT-4 (or your preferred model). Use a prompt that prompts GPT to return a strict JSON payload. Example prompt:

{"role": "user", "content": "Classify the inbound email into: priority (High, Medium, Low), category, and a brief action summary. Return a JSON object: {"priority": "High|Medium|Low", "category": "...", "summary": "..."}. Email subject: {subject}. Email body preview: {bodyPreview}."}

In Make, map the email fields into the prompt variables and ensure the GPT response is parsed as JSON. Validate that the JSON contains the expected keys and handle errors gracefully by sending a fallback log entry.

4. Slack module: add a Post a Message action. Choose your alert channel and configure the message to include dynamic fields from Gmail and the GPT response. Example message:

Priority: High From: {sender} Subject: {subject} Summary: {summary}

Include a compact, readable format and consider using a short attachment with a clickable thread link if available in Gmail.

5. Conditional router: add a filter after the GPT step. If JSON.priority equals High, route to Slack. If Medium or Low, skip Slack and proceed to logging only. This keeps noise out of your urgent channel while preserving full visibility in the log.

6. Google Sheets module (optional log): add Row action to append a record with timestamp, sender, subject, priority, category, and the GPT-provided summary. Include a column for the original message link if you can generate one, or a placeholder and a note to follow up.

7. Error handling: add a basic error path that, on any failure in OpenAI or Slack, posts a warning to a dedicated Slack channel (e.g., #ai-errors) and writes an error row to the log with the email subject and a fallback message.

8. Testing: send a test email to the Gmail inbox with known content that should produce High priority. Verify Slack alert appears and the log updates with correct data.

9. Activation: switch the scenario from Draft to On and monitor for the first 24–48 hours. Tune the prompt or routing rules based on observed results.

Real-World Business Scenario

A digital marketing consultancy receives dozens of inbound emails daily from clients and prospects. With this setup, urgent requests from key accounts are instantly surfaced in Slack for the account lead to respond within minutes. Meanwhile, all inquiries are logged for monthly reporting and to improve the classification prompt over time. The operations team reports a noticeable drop in triage time and a more consistent follow-up cadence, especially during peak weeks.

This pattern mirrors real-world integration work we’ve delivered for clients. If you want a version tailored to your environment, Olmec Dynamics can design, implement, and optimize it for you. Learn more about our automation work at Olmec Dynamics by visiting our case studies and services.

Common Variations

• Extend the log: push entries to a CRM or project management tool when a High priority email is detected, creating a task for the owner with a due date and context from the email.
• Generate a quick PDF summary: use a subsequent action to render a one-page brief from the subject and GPT summary and attach it to the log or a Notion page.
• Add user feedback: after Slack alert, prompt the owner with a quick reaction button to confirm whether the catgory and priority were correct, and train the model with corrections.

What this means for your team

This end-to-end automation gives your team a scalable way to triage emails with AI and act on critical items fast. It reduces cognitive load, improves response times, and creates an auditable trail for performance improvements. If you’d like this exact workflow adapted to your environment, Olmec Dynamics can help you design, implement, and optimize it. Learn more about our automation work at Olmec Dynamics by visiting our case studies and services.