Name the client and the brief comes back with every competitor claim quoted and linked to the page it came from, one proposition that beats a named rival claim, and a production spec table that separates each platform's hard limits from its recommendations.
Writing the creative brief is one of 700+ pre-built Flows in Juma. Name the client and the research happens first: the brand's own live claims, the claims four comparable competitors are making right now, and each platform's published specification page for the placements the campaign actually needs.
What comes back is a branded PDF creative brief plus an editable CSV asset matrix, one row per asset. The competitor set arrives with its selection rule printed above it, so nobody has to guess why those four brands and not four others. Every claim carries the full URL it was quoted from rather than a bare link word, and anything that exists only on a deep blog post gets dropped instead of cited. The three creative territories differ by mechanism and each one states what it does that the other two do not, plus the specific reason it could fail. The production table splits each platform's hard limit from its recommendation, because a 125-character truncation point printed under a heading that says maximum is how assets come back built wrong. Strategy, taste and judgment stay human: the brief names a person as the decision gate and says plainly that no asset ships on the brief alone.
Start with the claims, not the positioning. The Flow fetches the brand's homepage, product pages and about page, then does the same for four competitors chosen by a stated rule: same buyer, same price band, same sales channel. Every claim comes back quoted word for word with the full URL beside it and the date it was checked, and any claim carrying a dagger or asterisk gets a note explaining what that symbol obligates. The proposition is then written against that evidence and names the specific rival claim it beats. That is the difference between a brief that asserts whitespace and one that shows it.
Before writing anything, audit the live claims our client and its four closest competitors are making right now. State the selection rule first, then quote each claim verbatim with the full page URL and the date you checked it. Cite only homepages, product pages and about pages, never a deep blog post. Then write one proposition none of them is making, and name the claim it beats.
Most briefs come back with one idea wearing three coats. The fix is to make each route earn attention through a different mechanism and to say so out loud. One route might make evidence the experience, another might make the audience's own scepticism the hero, a third might dramatise the daily routine. Each territory carries a named insight, one hook line and the specific reason it could fail. A cliche check runs across the hooks and prints its result, so no route opens on a stock category phrase and no two hooks start with the same word.
Give me three creative territories that differ by mechanism, not by tone. For each one state in a single line what it does that the other two do not, then give the named insight, one hook line and the specific reason it could fail. Run a cliche check across the three hooks and print the result: no stock category phrases, and no two hooks opening with the same word.
More than a ratio. Each placement row carries the aspect ratio, a numeric safe zone, the file type, the duration limit and the platform's own specification URL. The character count is split into two columns: the hard limit the platform enforces and the point at which copy truncates or the platform simply recommends less. Where a platform publishes no safe-zone number, the row says so explicitly and gives the workaround rather than leaving a blank. Every placement in the table reconciles one to one against the CSV, and the count is printed with its arithmetic so the check is visible instead of asserted.
Build the per-placement production table for the placements this campaign needs. Split the character count into two columns: hard limit and recommended or truncates at. Give a numeric safe zone per row, or print "no published numeric safe zone" and say what to do instead. Cite the platform's own spec page in every row with the date you checked it, then reconcile the table against the asset matrix and show the arithmetic.
Regulated categories punish vague briefs. The do-not-say list comes back grounded in the category's own claim language, with each entry naming the rule it protects against rather than offering general caution. Disease claims, guaranteed outcomes, unsupported comparative safety language and undisclosed endorsements each get called out, and anything carrying a required disclaimer is flagged with what that disclaimer obligates. The point is not to make the creative timid. It is to let the team argue about the idea rather than discovering the problem after production.
Write the do-not-say list for this category. Each entry names the phrase to avoid and the specific rule it protects against. Cover disease and outcome claims, guaranteed results, comparative safety language and endorsement disclosure. Where a claim we quote carries a required disclaimer, say what that disclaimer obligates.
With a scored sheet written before anyone sees the work. The sign-off section lists the questions the reviewer scores each concept against, the scale they score on, and the threshold below which a concept cannot progress. Questions cover whether the proposition lands in the first beat, whether the work would still be recognisable with the logo removed, whether every factual assertion is specific and linked to real evidence, and whether a reasonable viewer could infer a claim nobody approved. The sheet is labelled as the human decision gate: it makes the judgment consistent, and a named person still makes it.
Write the reviewer sign-off sheet for this brief. List the scored questions we judge returned creative against, the scoring scale, and the score below which a concept cannot progress. Cover proposition clarity, brand distinctiveness, evidence quality, net impression and placement fit. Label it as the human decision gate and name who owns the final call.
One Project per client. Add the context once and every brief after the first starts from what the team already knows about this brand rather than from the live website alone. Juma remembers the approved claims, the visual system and the routes that have already run, so the second brief argues about the idea instead of re-establishing the basics.
Approved claims library
The claims legal has actually cleared, with the substantiation behind each one. Live site copy and the cleared list are not always the same thing, and this is what keeps the do-not-say section specific to this client.
Brand guidelines
Logo, colour, typography and the tone rules. With these in the project the PDF comes back looking like the client's own document rather than a generic template.
Past campaign briefs and what got made
The routes that have already run and how they performed. This is what stops a new brief proposing the territory the team shot last quarter.
Each knowledge item takes a one-line description telling Juma what it is for. A short, plain sentence is enough.
A creative brief is the document a marketer gives a designer, editor or agency before any asset gets made. It fixes the audience, the single proposition the work has to land, the routes worth exploring, the language that cannot be used, and the technical shape of every asset the campaign needs.
It is not the ad and it is not the media plan. It is the decision record that keeps three people making three different things from arriving at three unrelated answers, and it is the thing a reviewer scores returned work against.
This Flow stops before the ad. It produces the brief that tells someone else what to make and how it will be judged, which is a different job from writing the finished headline.
If the team wants the ads themselves, the copy Flows cover that ground: Google Ads copy, Facebook ads and LinkedIn ad copy each produce finished variants. The brief is the upstream document those executions get measured against, and it is the artifact the external agency actually needs.
Yes. Each placement row carries the aspect ratio, a numeric safe zone, file type, duration limit and the character counts, quoting the platform's own specification page with the date it was checked.
The character count is deliberately split into the hard limit and the recommendation, because those two numbers get confused constantly and the confusion is expensive: copy built to a recommendation printed as a maximum gets truncated, and copy built past a real limit gets rejected. Specifications also change, so the checked date is printed on every row and the brief tells the team to confirm in the platform before trafficking.
Both. The structure is the same either way: the claims your competitors make, one proposition that beats a named claim, three routes, the specs and the guardrails.
What changes is the evidence and the placements. A consumer supplement campaign pulls product-page claims and Meta placements. A B2B software campaign pulls security and compliance claims and LinkedIn placements, with the regulated-language section shifting from health claims to whatever the category actually constrains.
A person, always. The brief carries a reviewer sign-off sheet with scored questions and a threshold, and it names the human decision gate explicitly.
Juma augments the team, it does not replace it. The Flow does the research, the sourcing and the specification work that eats an afternoon, so the argument that matters, which route to make, happens with better evidence and with a named person still making the call.
Name the client, the campaign, and the audience. Juma writes five Facebook ad copy variants with distinct hook angles, validates every character count against Facebook's limits, and delivers an Excel file ready to upload to Ads Manager.
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
Start typing a task and Juma will figure out the rest.
.webp){kind=icon}