How we turned campaign and client briefs into an AI creative engine and content calendar system

Every campaign started the same way: a brand profile and a campaign brief, both read by hand, then manually translated into a content plan before a single post could be written. Doing that translation consistently, across seven different social platforms, was the actual bottleneck.

Each platform needed its own tone, format, and structure. A LinkedIn post and a TikTok script share almost nothing structurally, so writing for all seven from the same brief meant rewriting the same idea seven separate times, often losing the thread of the original brand tone along the way.

What the team needed wasn’t a faster writer. It needed a system that could read the source documents once, plan a content structure grounded in what those documents actually said, and then generate platform-correct copy for every platform in scope without a person rewriting each one by hand.

We split the build into two linked systems rather than one large tool: a planning stage that turns source documents into a content structure, and an execution stage that turns that structure into finished post copy, each independently testable and reusable on its own.

The planning stage starts from an intake capturing the brand profile document, the campaign brief, a start and end window, and a target post volume. It reads both source documents and maps them into a fixed brand profile, name, industry, tone of voice, up to four content pillars, and stated dos and don’ts, instructed never to invent a brand fact and to explicitly log any assumption it does make.

From that grounded profile, a second step builds the actual content structure: a set of pillars and milestones forming the campaign’s shape, then expands every milestone across each of the seven platforms in scope, Instagram, Facebook, LinkedIn, X, Threads, YouTube Shorts, and TikTok, each with its own reference ID, before writing the full structure into a spreadsheet as a placeholder awaiting finished copy.

The execution stage picks up from there: rather than generating a post for every single pillar-milestone-platform combination, it samples evenly across pillars and platforms up to the requested post volume, so output stays reviewable instead of flooding the calendar with content for every theoretical combination.

Each sampled post is generated against a platform-specific style guide, tone, length, hook style, and call-to-action conventions unique to that platform, then parsed, matched back to its original pillar and milestone, and written directly into the shared spreadsheet, so the campaign’s content calendar updates in place rather than duplicating anywhere.