AI automation tutorials for beginners

Last month I needed to enrich a list of 500 leads scraped from LinkedIn with company size and industry, then send a personalized intro email. I had ChatGPT plus a Google Sheet but no way to connect them without manual copy‑paste. After a few hours of trial, I built a simple automation that pulls each row, asks GPT to classify the company, writes a first line, and pushes the result back to the sheet.

What most guides get wrong

Most beginner tutorials start by recommending a heavyweight platform like Zapier or Zoho Flow and then drown you in pre‑built templates that hide the underlying logic. They assume you want a click‑and‑play solution, but the real value comes from understanding how data moves between steps. If you skip that understanding, you’ll be stuck when the template breaks or when you need a custom step that isn’t in the library.

Instead, start with the most minimal tool that lets you see each action and its output. For this guide we’ll use Make (formerly Integromat) because its visual scenario builder shows every module and the data passing between them, which makes debugging obvious.

How to pick a no‑code AI platform that won’t lock you in

Is the free tier enough to test a simple workflow? Yes, if you stay under 1,000 operations a month. Make.com’s free plan gives you 1,000 operations, which is plenty for a prototype that runs a few dozen leads each day. If you need more, the $29/mo “Core” plan bumps you to 10,000 operations and adds multi‑step error handling.

I think the $29/mo price is fair for a solo operator who wants reliable scheduling and decent logs, but the jump to the $99/mo “Team” plan feels steep unless you’re running dozens of scenarios simultaneously. Always start free, then upgrade only when you hit the operation limit.

Building the enrichment workflow step by step

Here’s the exact sequence I used, with the prompts and settings that worked.

1. Create a new scenario in Make.com.
2. Add a “Google Sheets – Watch Changes” module set to your lead spreadsheet. Choose the sheet and the range that holds the raw LinkedIn data.
3. Add an “HTTP – Make a request” module to call the OpenAI Chat Completions endpoint. Use your API key, set the model to gpt-3.5-turbo, and paste the following prompt into the message body:
   
You are a data enrichment assistant. Given a company name, return a JSON object with two keys: "size" (choose from "1-10", "11-50", "51-200", "201-1000", "1000+") and "industry" (pick one from the list: SaaS, E‑commerce, Manufacturing, Healthcare, Finance, Education, Other). If you are unsure, set both values to "unknown".
Company: {{1.value}}

   (Here {{1.value}} is the output from the Google Sheets module containing the company name.)

4. Add a “JSON – Parse” module to turn the OpenAI response into usable fields.
5. Add a second “Google Sheets – Update a row” module. Map the parsed size and industry to two new columns in the same row.
6. Optionally, add a “SMTP – Send an email” module that uses the enriched data to craft a first line: “Hi {{2.value}}, I noticed {{1.value}} works in the {{3.value}} space…” (Replace {{2.value}} with the first name column, {{3.value}} with the industry.)
7. Save and run the scenario once manually to verify each step. Check the operation log in Make.com to see the data moving between modules.

When I ran this on 500 rows, it consumed about 1,200 operations (one watch trigger, one OpenAI call, one update per row). That stayed well within the free tier limit for a single day’s batch.

How to debug when this breaks

The most common failure point is the OpenAI module returning a non‑JSON response, which kills the parser and stops the scenario. If you see a red error bubble on the HTTP module, click it to view the raw response. Often the model adds extra commentary before or after the JSON. To fix that, wrap the prompt in a system message that insists on JSON only, or add a “Text – Replace” module to strip anything that isn’t curly braces before the parser.

Another frequent issue is the Google Sheets watch module missing changes because the spreadsheet is set to a different timezone than your Make.com scenario. Make sure both are set to UTC, or adjust the watch module to poll every five minutes instead of relying on instant triggers.

Finally, if you hit the operation limit, Make.com will pause the scenario and send an email warning. The log will show “Quota exceeded” on the module that tried to run. Either wait until the next billing cycle or upgrade the plan.

Price check: what you actually pay to run this at scale

Running the enrichment on 5,000 leads a month costs roughly:

• Make.com Core plan: $29/mo (10,000 operations)
• OpenAI API: ~ $0.002 per 1,000 tokens with gpt-3.5-turbo. Each call uses about 150 tokens, so 5,000 calls ≈ $1.50.
• Google Sheets: free.

That’s under $31/mo total. I think that’s a reasonable price for a fully automated lead enrichment pipeline that would otherwise take a virtual assistant several hours each week.

One concrete gripe I have with Make.com is that the error log viewer shows only the most recent 100 lines and offers no search or export option. When a scenario fails repeatedly, I have to copy‑paste the log into a text editor just to grep for a specific token. It’s annoying, but the visual builder more than makes up for it.

One concrete love is the ability to pause a scenario at any module and inspect the exact data payload that’s waiting to move forward. That single feature has saved me countless hours of guesswork when a step behaves unexpectedly.

For more on this exact angle, AI meeting tools coverage.

If you’d rather skip the build and deploy a working version in an afternoon, we’ve packaged this workflow as a blueprint at deepusecase.com/vault/ai-automation-blueprint.