Automating content briefs: what the brain adds that a template cannot

Content briefs sit in the middle of our ranking of what to automate for a specific reason: templates already exist, so the question is not whether briefs can be systematised — every agency has a template — but what an agency brain adds that a form does not. The answer is memory, and it changes what a brief is for.

What a template knows

A brief template knows the shape of a good brief: target keyword, search intent, suggested headings, word count, internal links, who the reader is. Fill it in well and you get a competent commission. The problem is who fills it in. The template is empty until a strategist spends forty minutes gathering what the template asks for — and most of that gathering is not thinking, it is retrieval from places the strategist has to go and look.

What the brain knows

A brief drafted from an agency brain starts from three kinds of memory a template cannot hold.

What this client has already published. Not a vague sense of it — the actual list, with performance attached. The drafted brief does not propose a topic the client covered fourteen months ago, and when a new piece should reference or update an old one, it says which.

What this client rejected, and why. Every account accumulates a history of angles that were pitched and declined — too salesy, too technical, legal did not like it. In most agencies that history lives in whoever has been on the account longest, which is why a new strategist re-pitches a dead angle in their first month. A brain that has read the feedback threads does not.

The house voice, from evidence. Style guides describe voice in adjectives nobody can apply. Two hundred published pieces describe it by example. A brief that carries “this client writes like this, not like that” — with real excerpts — gets a writer to the right register in one draft instead of three.

The practical output is a brief a writer can start from, rather than a form a strategist has to fill in. The strategist’s forty minutes of gathering drops to a review pass; the writer’s first draft lands closer. As with reporting, the saving is concentrated in the assembly, which was most of the hours and none of the value.

What stays human

Deciding what is worth writing about at all. A brain can surface gaps and suggest topics from search data, and the suggestions are useful inputs — but the call about what serves this client’s strategy this quarter is judgement with commercial consequences, and it stays with a person for the reasons we hold everywhere.

So does the final read of the brief itself. A wrong brief is the most expensive document in the content process, because every error in it is multiplied by the writer, the editor and the revision cycle that follows.

The drift to watch for

A brief drafted from history is a brief anchored to history. Run this for six months without supervision and there is a version where every proposed topic is adjacent to an existing topic, every angle resembles a previous angle, and the client’s content quietly converges on itself. The model is doing exactly what it was asked — drafting from what exists — and the result is strategic stagnation with excellent consistency.

The fix is a standing habit, not a technical one: someone reads a month of briefs in one sitting, quarterly, and asks a single question — would anything on this list have surprised us a year ago? If the answer is no three quarters running, the human layer has stopped doing its job, not the machine.

How to start

Do not automate the whole brief first. Automate the research appendix — the published-list, the performance pull, the excerpts — and keep writing the strategic half by hand on top of it. You get most of the hours back immediately, you learn what the brain gets wrong while it is still harmless, and the strategists experience the system as an assistant before they are asked to trust it as a drafter. Two months of that and the right scope for full drafting will be obvious from your own evidence rather than our claims.