How we run our own studio

The situation

We take busywork off founders' plates and then automate it, so our own studio runs the same way. The blog, the social posts, the free ops audit and the search for new work are each AI plus code checks plus a person who approves before anything goes out. Most of what these systems do is say no: below are the real counts from their first days, including what they rejected.

What we built

• A blog writer that runs three times a day. It picks a topic by type and reader, drafts from a weekly list of checked sources and our own field notes, then has to pass code checks (length, links, banned phrases) and a separate editor pass that scores six things. A draft gets two rewrites to reach the bar, or it is skipped.
• Social carousels and captions made from each published guide. A fact check reads every line against the guide and our field notes. Nothing goes to Facebook, Instagram or LinkedIn until a reviewer approves it on our admin page.
• A weekly sources scan that collects reports and keeps a number only if it appears on the page it is cited from.
• The free ops audit at /audit. The first recommendation shows on the page, the request is saved to our database, and a job on our side drafts the full audit as an email for a person to check and send. It never sends on its own.
• A job scan that reads public freelance posts on seven sites, scores each one against our packages and drafts a proposal only when a post clears the bar. A person submits it.