You can build a working AI inventory alert system in under an hour using Make (formerly Integromat), OpenAI’s API, and Slack—no developer needed, no custom code, and it will genuinely save your ops team from babysitting spreadsheets every morning. This tutorial walks through the exact setup, the gotchas I hit, and what this system can and can’t do for you.
Why Bother With AI-Powered Inventory Alerts?
Basic threshold alerts already exist in most inventory tools. Shopify, for instance, lets you set a reorder point and will ping you when stock hits it. The problem? That ping is dumb. It tells you “Widget A is at 12 units” and nothing else. You still have to open four tabs, check your sales velocity, look at your supplier lead times, and decide whether 12 units means you’re fine for two weeks or you’re about to blow a product launch.
The AI layer changes that. Instead of getting a raw number, you get a message that says: “Widget A is at 12 units. At your current 7-day sales velocity of 4 units/day, you have roughly 3 days of stock left. Your supplier lead time is 8 days. Recommend placing a PO today.” That’s actionable. That’s the difference between an alert and a decision.
I set this up for a small consumer goods brand doing about $1.2M/year in Shopify revenue. Before the system, their ops manager was spending 45 minutes every morning reviewing inventory in a Google Sheet. After? She checks Slack once, acts on the alerts that matter, and moves on. That 45 minutes dropped to under 10.
What You Need Before You Start
Here’s the stack and the honest cost breakdown:
- Make (formerly Integromat) — The automation backbone. The Core plan runs $9/month and gives you 10,000 operations/month, which is more than enough for most small businesses running this workflow. Make is genuinely more flexible than Zapier for multi-step logic, though its interface has a steeper learning curve. If you want context on how these platforms compare, this breakdown of Zapier vs. Make vs. n8n is worth reading first.
- OpenAI API — You’ll use GPT-4o-mini for cost efficiency. At roughly $0.15 per 1 million input tokens, running 50 inventory checks per day costs you pennies. Literally. Budget $2–5/month for this unless you’re a massive retailer.
- Slack — Free tier works fine. You’re just sending messages to a channel.
- Your inventory data source — This tutorial uses a Google Sheet, because it’s the lowest common denominator. Shopify, WooCommerce, or a direct database connection all work through Make’s native integrations.
You also need an OpenAI account with API access enabled and billing set up. Takes five minutes at platform.openai.com.
Step 1: Structure Your Inventory Google Sheet
Your sheet needs specific columns for the AI to have anything useful to say. Here’s the exact structure I use:
- Column A: SKU
- Column B: Product Name
- Column C: Current Stock (units)
- Column D: Reorder Point (units)
- Column E: 7-Day Sales Velocity (units sold in last 7 days)
- Column F: Supplier Lead Time (days)
- Column G: Preferred Reorder Quantity
Columns E and F are where most people get lazy. Don’t. The AI’s analysis is only as good as the data you feed it. If you’re on Shopify, you can pull 7-day sales velocity with a simple report export and paste it in weekly, or automate that pull with another Make scenario. Column F should reflect your actual supplier lead times—not the optimistic number your vendor quotes you, the real one.
Step 2: Build the Make Scenario
Open Make and create a new scenario. Here’s the module sequence:
Module 1: Schedule Trigger
Set this to run every morning at 7:00 AM in your timezone. Use Make’s built-in Schedule module. No configuration beyond setting the time and selecting “Every Day.”
Module 2: Google Sheets — Search Rows
Connect your Google account and select your inventory sheet. In the Filter section, add a condition: Column C (Current Stock) is less than or equal to Column D (Reorder Point). This means Make only pulls rows where you’re at or below reorder threshold—not your entire catalog. If you have 400 SKUs and 12 are low, you’re sending 12 rows to the AI, not 400. Keep your costs and noise level down.
Module 3: Iterator
Drop an Iterator module after the Google Sheets search. This tells Make to process each low-stock row individually rather than as a batch. Critical step—skip it and your AI call gets messy.
Module 4: OpenAI — Create a Completion
This is the core of the whole thing. Connect your OpenAI API key. Select model gpt-4o-mini. In the prompt field, use a system message plus a dynamic user message. Here’s the exact system message I use:
“You are an inventory analyst for a small retail business. When given inventory data, you write a concise, plain-English Slack alert (2–3 sentences max) that tells the ops team what the situation is, how urgent it is, and what action to take. Be direct. No fluff.”
Then the user message pulls in Make’s dynamic variables from the iterator:
“Product: {{2.Product Name}}. SKU: {{2.SKU}}. Current stock: {{2.Current Stock}} units. Reorder point: {{2.Reorder Point}} units. 7-day sales velocity: {{2.7-Day Sales Velocity}} units. Supplier lead time: {{2.Supplier Lead Time}} days. Preferred reorder quantity: {{2.Preferred Reorder Quantity}} units.”
Set max tokens to 150. You don’t need a novel—you need a decision-support message.
Module 5: Slack — Create a Message
Connect Slack and pick your #inventory-alerts channel (create it if it doesn’t exist). In the message body, pull in the OpenAI response: {{4.choices[].message.content}}. You can also add a header line before it—something like “⚠️ Low Stock Alert: {{2.Product Name}}”—so the channel stays readable at a glance.
Step 3: Test It on Real Data
Before you let it run on a schedule, drop a product below its reorder point in your sheet manually and click “Run Once” in Make. Watch each module light up. Check that the Google Sheets module catches the row, the OpenAI module returns a real message (not an error), and Slack receives it cleanly.
The most common failure point here is the OpenAI API key. Make sure it has billing enabled and isn’t using an old key from a project that was deleted. I’ve been burned by this twice.
Also watch the token count on that OpenAI call. With gpt-4o-mini at 150 max tokens, you’re well within limits, and each call costs roughly $0.000023. If you have 20 alerts firing daily, that’s less than $0.0005/day. Genuinely negligible.
Step 4: Add a Weekly Summary Layer (Optional but Worth It)
Daily alerts handle urgency. But I also recommend a separate Make scenario that runs every Monday morning and sends a broader weekly inventory health summary. This one pulls all SKUs—not just low-stock ones—computes days-of-stock-remaining for each (Current Stock ÷ 7-Day Velocity), and sends a ranked list to Slack sorted by most urgent to least.
For this version, I switch to gpt-4o rather than mini, because I’m asking it to reason across a larger dataset and produce a structured ranked summary. The cost bump is still trivial—maybe $0.03 per weekly run—but the output quality difference for complex multi-item reasoning is noticeable. If you’re curious about how these AI-to-workflow connections work at a deeper level, this guide on AI workflow orchestration gives good conceptual grounding.
Honest Limitations You Should Know
This system is not magic. The AI isn’t actually “watching” your inventory—Make is triggering it on a schedule, so your alerts are only as fresh as your last run. If you sell out of something at 2 PM and your scenario runs at 7 AM, you won’t know until tomorrow morning unless you also set up a second trigger (say, every 4 hours during business hours).
The AI recommendations are probabilistic, not authoritative. If your velocity data is stale or your lead time column hasn’t been updated since your supplier changed their shipping policy, the AI will give you confidently wrong advice. Garbage in, garbage out—the AI just makes the garbage sound professional.
Make’s Google Sheets integration also has a known quirk: it reads sheets row-by-row and can sometimes time out on very large sheets (500+ rows) on the Core plan. If you have a massive catalog, filter your sheet to a summary tab before Make reads it, or consider upgrading to Make’s Pro plan at $16/month.
This kind of automation compounds nicely with other operational improvements. If you’re already thinking about broader admin automation, this case study on an e-commerce brand that eliminated 14 hours of weekly admin shows what’s possible when you stack multiple workflows together.
My Actual Recommendation
Start with the daily alert scenario only. Get it stable, run it for two weeks, and tune your reorder points and velocity data until the alerts feel right—not too noisy, not missing real problems. Then add the weekly summary. Don’t try to build both on day one; you’ll spend more time debugging than you’ll save in the first month.
Use gpt-4o-mini for daily alerts (cost-efficient, fast, perfectly adequate) and only consider gpt-4o if you add the multi-SKU summary layer. Don’t use GPT-4 Turbo for this—overkill by a wide margin, and the cost difference isn’t justified for what amounts to filling a template with business logic.
If you’re also looking to automate other repetitive business outputs, the same Make + OpenAI pattern applies surprisingly well to things like proposal drafts. Here’s how one team cut proposal time from 4 hours to 22 minutes using a nearly identical approach.
FAQ
Do I need coding skills to build this inventory alert system?
No. Make’s interface is visual—you’re connecting modules, not writing code. The only text you write is the AI prompt, and the examples above give you exactly what to use. If you can use Google Sheets and Slack, you can build this.
What if my inventory data lives in Shopify or WooCommerce instead of Google Sheets?
Make has native integrations for both platforms. Swap the Google Sheets module in Step 2 for Make’s Shopify or WooCommerce “Search Products” module, filtering by inventory quantity below your threshold. The rest of the scenario stays identical. Shopify’s API also returns sales data directly, so you can calculate velocity inside Make rather than maintaining a separate sheet column.
How do I keep the AI from sending alerts for products I’ve intentionally run down?
Add a column to your sheet called “Alert Active” with a simple Yes/No value. In Make’s Google Sheets filter, add a second condition: Column H (Alert Active) equals “Yes.” Products you’re discontinuing or intentionally depleting get set to “No” and drop out of the alert queue entirely. Takes 30 seconds to add and saves you from alert fatigue immediately.
This article was produced with the assistance of AI, and its featured image was AI-generated. We review for accuracy, but please verify critical details.
