The repetitive part of monthly client audits is not the analysis — it is the transcription. Looking at a Search Console screenshot and typing "clicks: 271, impressions: 180,234, CTR: 0.2%, average position: 28.7" into a report template. Twenty times. Every month. Across five clients.
I am building an n8n + Python workflow that watches a monthly audit folder, classifies new screenshots by platform, extracts visible metrics, compares against last month, and feeds the structured data to a Python DOCX generator. The output is a draft report in the correct client folder.
The human checkpoint
The pipeline ends with a draft, not a published report. Every audit gets a manual review pass before delivery — because SEO metrics, screenshots, and AI interpretations all need a human reading them in context. Automation cuts the boring 80%. The interesting 20% stays mine.