My Complete AI Content Workflow: From Brief to Published in 90 Minutes

When I tell people I can go from a client brief to 30 published-ready ad variants in under 90 minutes, they assume I'm cutting corners. I'm not. I've built a system that handles the mechanical execution while I focus on the creative judgment that AI can't replicate.

The Stack

• Claude 3.5 Sonnet: All copywriting, angle development, strategic analysis
• n8n (self-hosted): Orchestration layer connecting everything
• Midjourney + Ideogram: Image generation — Midjourney for photo-realistic, Ideogram for text-heavy Arabic designs
• ElevenLabs: Voice-over scripts for video ads
• Google Sheets: Single source of truth for briefs and approvals

The Workflow: Step by Step

1. Brief input (8 min): Coordinator fills Google Sheet — product, audience, campaign goal, key features, tone restrictions
2. n8n trigger: Webhook fires when row status = "APPROVED"
3. Research sweep: Perplexity API pulls current market context, competitor positioning, trending angles for Kuwait/GCC
4. Strategic round with Claude: Research + brief → 5 distinct strategic angles in JSON format, each with hook, emotional driver, CTA direction
5. Human gate (me): Approve 2-3 angles via Telegram reply
6. Full creative generation: 10 copy variants per approved angle — 3 headlines, 3 body copies, 2 CTAs, 2 Arabic adaptations
7. Image prompts: Claude writes Midjourney prompts → n8n sends to API → stores in Drive
8. Airtable assembly: Copy + images assembled into ready-to-execute records

The Key Insight

I didn't automate creativity. I automated the mechanical work surrounding creativity: research compilation, brief formatting, first-draft generation, variant production, and report assembly. Creative judgment stays human.

Real Results After 6 Months

• Average ROAS: 2.1x → 4.5x (faster creative iteration = more winning ads)
• Time from brief to published: 5 business days → 18 hours average
• Creative variants tested per month: 4-6 → 18-22
• Team cost: reduced ~55% year-over-year

"The biggest mistake in AI content generation is treating the model like a vague employee. Treat it like a talented contractor who has never met your client — every constraint, preference, and 'we don't say this' must live in the system prompt."