AI Workflow AutomationCase study

80+ clients. Fifteen posts a month each. One person reviewing.

Three hours a day on Google Business Profile content became two hours a week. The agency stopped producing and started doing strategy.

Built with
n8nClaudeAirtable
G
Local business
Open now
4.8 · 142 reviews
Posted 5 hrs ago by AI
n8nClaudeAirtable
80+ profiles

80+

Profiles

15

Posts / client / mo

3 hrs

Daily → weekly

1

Reviewer

01 / What was happening

Three hours a day. Generic posts. One bottleneck.

A digital marketing agency managed Google Business Profiles for over 80 clients. Each client needed 15 posts a month. Each post needed to reflect that specific business, in that specific market, at that specific time of year. A plumber in Tampa in December gets different content than an accountant in Phoenix in April.

The SEO person responsible was spending three hours a day on GBP work alone. Quality was inconsistent. Posts were generic. The people who should have been doing strategy were trapped in production.

Workload · daily
3 hrs · 80 profiles · 1 person
02 / What we built

Per-client intelligence. One dashboard. Scheduled output.

An AI content system using n8n and Airtable. Built around the insight that generic prompts produce generic content. The difference is the research underneath each post. Three layers, one dashboard.

01 · Per-client intelligence
n8n
Claude

Research per business

The system knows each business: their type, their market, their promotions, what's seasonally relevant. A post for the plumber doesn't read like a post for the accountant, because the research behind each is different.

02 · Central dashboard
Airtable

Single review surface

One Airtable base. One view per client. Review, edit, and approve from a single screen. Batch approval for the ones ready. Individual editing for the ones that need work.

03 · Scheduled publishing
n8n

Auto-publish · GBP

Approved posts go live on schedule. No manual intervention. Across 80+ profiles, with seasonal calibration baked in.

03 / What changed

The team stopped producing. Started strategising.

  • 80+profiles managed through one system.
  • 15posts per month per company, generated and published.
  • 3 hrs/day → 2 hrs/wkon GBP content. The math speaks.
  • the team stopped producing and started doing the strategic work they were hired for.
Editorial dashboard · this week
Tampa Plumber · holiday rate noticeapproved
Phoenix Accountant · Q2 deadlinesapproved
Atlanta Roofer · storm season prepqueued
Boise Dentist · new whitening svcqueued
Reno HVAC · summer tune-upapproved
Charlotte Lawyer · estate planningdraft
Denver Salon · spring coloursqueued
Approve · Schedule

05 / The lesson

Generic in, generic out. Specific in, specific out.

The model didn't change. The research did. A generic prompt with a business name produces generic content. A prompt with three layers of context produces content a human would be proud to post.

The first version pulled business type and location. Posts were technically correct but flat. When we rebuilt the research layer to include seasonal trends, recent reviews, and competitor activity, the posts started reading like someone who actually knew the business.

This was our first automation project. The proof of concept that showed AI content systems could replace manual production without the output getting worse.