n8n
Claude
Google Sheets
Meta Ads
Klaviyo
The problem
Every brief followed the same pattern. The copywriter opens Sheets, searches for relevant past briefs manually, checks Meta Ads for what performed, guesses at brand voice from examples they remember, then writes one or two variations. That's 45 to 60 minutes of setup before any actual writing starts, every single brief.
All the inputs this process needs already exist in structured systems. Briefs are in Google Sheets. Performance data is in Meta and Klaviyo. Brand voice can be extracted from the best-performing approved briefs. The workflow just didn't exist to pull them together.
How it works
The workflow runs on a schedule. No form to fill, no trigger from a human. It wakes up, collects everything it needs, generates the copy, and drops it where the team expects it.
The AI prompt combines the current brief with the top-performing approved examples sorted by engagement score, plus a live performance summary from Meta and Klaviyo. The model sees what good looks like for this specific brand, what's been performing numerically, then the actual task. Output is structured JSON so each variation writes back cleanly to the right Sheets column.
What changed
Copywriter manually reviews past briefs, checks ad manager for recent performance, writes 1–2 variations from scratch. 45–90 minutes of prep per brief, inconsistent brand voice across writers.
Brief is flagged ready. Workflow runs overnight. Writer opens Sheets in the morning to find 3 on-brand variations with performance context attached. Review takes 10 minutes.
Every variation is written with context from the team's top-performing Meta and Klaviyo campaigns — not generic best practices.
Approved copy feeds back into the reference pool. The model's understanding of brand voice improves with every brief that goes through.
Consistency is the less obvious win. When brand voice lives in a system instead of individuals, it doesn't drift when someone is on holiday or a new hire joins.