How I Use Claude Code to Automate My Ad Reporting

This is an example article so you can see how a full post looks. Replace it with your own content whenever you’re ready.

For years, Monday morning meant the same thing: export the numbers, wrestle them into a spreadsheet, format the charts, write the summary. Half a day, gone — before I’d made a single decision.

Then I moved the boring parts to Claude Code. Here’s how it actually works, and where I still keep a human in the loop.

What AI is genuinely good at here

Not “insights”. Not strategy. The plumbing:

• Cleaning and joining raw exports from Meta and Google.
• Recalculating the same metrics the same way, every single week.
• Formatting a tidy report and drafting a first-pass summary of what changed.

That’s repetitive, rule-based work — exactly where AI shines and humans make silly mistakes.

My setup in three steps

1. Drop the exports in a folder. Same file names every week (naming discipline again — it always comes back to that).
2. One command. Claude Code runs a small script that cleans the data, calculates spend, ROAS and CPA per campaign, and builds the report.
3. I review, then decide. The AI hands me a clean draft. I bring the judgement.

Where AI still needs a human

AI will happily tell you a campaign “improved” because ROAS went up — while ignoring that spend collapsed and you missed your revenue target. It has no idea what the business is trying to do.

So the rule in my accounts is simple:

Let AI do the counting. You do the deciding.

Should you try it?

If you’re copying numbers between tabs every week, yes. Start with the single most repetitive report you make, hand the mechanical part to Claude Code, and keep the judgement for yourself. You’ll get the afternoon back.