AI Content & Reporting Pipeline
The situation
The situation
A digital marketing agency was managing SEO content production for 20+ clients — each with their own tone, keyword focus, and monthly reporting requirements. Every article was written by a human writer, briefed manually, edited, and uploaded. Monthly reports were assembled by pulling data from Google Search Console, formatting it into a template, and writing a narrative summary. With 20 clients, this was consuming the majority of two people's time every month.
The complication
The complication
The agency's clients had explicitly asked for more content — they wanted 6–8 articles a month rather than 2–3. The agency couldn't quote for that volume without hiring more writers, and the economics didn't work at the rates clients were paying. They needed to produce more without proportionally increasing headcount.
What we built
The system, component by component
A keyword brief generator that pulls ranking data from Google Search Console via API, identifies content gaps, and produces structured briefs for each client automatically
A long-form article drafting pipeline using GPT-4o, with a per-client "voice profile" stored in Airtable that conditions the model on each client's tone, preferred structure, and banned phrases
An editorial review step where each draft is flagged for human review only if it falls below a quality threshold — most articles pass without changes
A monthly report generator that pulls GSC data, compares it to the previous period, writes a plain-English narrative summary, and formats it into the agency's report template as a ready-to-send PDF
An Airtable dashboard that tracks content status, publication dates, and report delivery for all 20 clients
Under the hood
Technical note
The voice profiles were built by having GPT-4o analyse each client's existing content and extract stylistic patterns — sentence length, formality level, topic framing. These are stored as structured prompts in Airtable and prepended to every generation request for that client. The result is that articles for a B2B SaaS client read noticeably differently from articles for a consumer lifestyle client — without manual intervention.
Stack used
What changed
What changed
The agency now produces 6–8 articles per client per month instead of 2–3, at the same cost base. The reporting that used to take two days of manual work each month now takes a few hours of review. A new client contract worth $4.2k/month was won specifically because the agency could credibly promise volume they previously couldn't deliver.
Next case study
Automated Lead Recovery System
Recognise any of this in your own business?
If you have a process that's eating time every week, tell us about it. We'll come back within 24 hours with an honest assessment.