A law firm partner's LinkedIn runs itself now.
6 hours a week of content creation replaced by an automated research-to-post pipeline. The partner reviews for five minutes. The system handles the rest.
Drafted by Claude · voice: Notion brand book
~20
Posts / month
30 min
Weekly review
6 hrs
Hours back / week
1 partner
Five-minute approval
Six hours a week. Or zero posts a month.
The managing partner of a law firm needed a LinkedIn presence. Thought leadership, industry commentary, the kind of posts that build credibility over time. But producing that content took six hours a week. Either the partner wrote it and had no bandwidth, or a team member handled it and had other priorities.
Posts were inconsistent. Three in a week, then nothing for a month. The agency managing their marketing needed a system that worked for this client and could scale across their portfolio.
Three layers. n8n underneath. Airtable on top.
A three-layer content pipeline running on n8n and Airtable. Each layer has one job. The handoff between them is what makes the system feel finished, not stitched.
Monitor · Filter · Flag
The system monitors the sources the partner actually reads. Industry publications, regulatory updates, competitor moves. Filters for relevance. Flags what's worth responding to.
Drafted in voice
Using that research as context, the AI drafts posts in the partner's voice and perspective. Not generic thought leadership. Posts grounded in specific, current developments their audience cares about.
Review · Edit · Schedule
Every draft lands in an Airtable dashboard. The agency reviews, edits, approves, schedules. Nothing goes live without a human saying yes.
Five-minute reviews. Twenty posts a month. Same partner.
- ~20posts a month produced automatically.
- 6 hrsper week freed up for the partner and agency team.
- 5 minpartner review window. The system handles the rest.
- 1 → Ndesigned for one client. Built to scale to every executive the agency manages.
05 / The lesson
Voice was the hard part.
LinkedIn is full of AI content that reads like AI content. Flat, safe, interchangeable. For this system to work, every post had to sound like the partner wrote it.
The build was straightforward. Connecting APIs, setting up Airtable, writing the n8n workflows. None of that was the hard part.
That meant building prompts with enough context about the partner's opinions, tone, and priorities that the AI couldn't fall back on generic output. The technical build doesn't need to be complicated. It needs to not fail. The value lives in the details that make the output indistinguishable from human work.
Next step