Writing a marketing brief with AI

A marketing brief is the one document that says who you are for, what you say to them, and where you spend your time. Everything else, the ads, the homepage hero, the posts, hangs off it. Without one you end up rewriting your positioning in every channel and quietly saying something different each time.

The reason most founders never have one is the same reason most never film their build: the blank page. You are asked to describe your buyer, your competitors and your message in the abstract, before you have the words for any of it. So it gets postponed, and the marketing happens on instinct instead.

Marketing brief sounds boring, but not an Underway one. The point is that the brief is not something you fill in. It is written from what you have already done: the site you published, the changes you shipped, the things you said out loud in your episodes.

That is the real shift with AI here. A general chatbot can write a brief if you describe your business to it well. The quality of that brief is capped by how well you described yourself on a day you could not be bothered. A brief written from your tools starts from evidence.

Here is how I do it. I go to the Promote tab, and at the top is your marketing brief. I click into it and press Write It. I had already done mine once, but I ran it again to work through the whole process on camera.

What it tells you up front is plain: Underway reads your site, what you have built and your episodes, looks up what your buyers use instead, then writes. About four minutes. You can leave this page and it carries on. So if you get bored, walk away from the screen. There is a loading mechanism at the top so you can see what it is doing while it runs.

That progress strip is worth watching once, because it is the proof that it is reading rather than guessing. It reads the webpage. It sees that I post on X and TikTok, because I am connected to them in the Context tab at the top. It read 537 changes I have shipped, which is GitHub. It read what I said in 10 episodes, for how I sound, so it can develop my voice for my marketing. It read 18 searches for how I show up on Google, zero clicks between them, because we have just started. It read five pages on my site. Then it goes and looks at what my buyers use instead.

Connections matter more than prompts here. The more of your working day is connected, the less the brief is inference. In Underway the connections are read-only: they shape the questions and the writing, they never change anything in your accounts.

When it lands you get nine sections: who it is for, positioning, messaging, where to spend the effort, your next 30 days, who else they look at, how you sound, the product, what it could not tell, and what it read.

That last pair are the honest ones and the ones to read first. What it could not tell is the list of gaps, the questions it cannot answer from your data and you have to decide. Mine includes whether the first message should lead on loss, nobody filmed it, or on no editing; whether the priority now is paid acquisition or earning the first paying customers from the existing audience; and whether the anyone making anything line belongs on the homepage now or stays a secondary page. Those are genuinely my calls, not the AI's.

What it read is the audit trail. It is how you judge whether to trust the rest. If the brief read five pages and ten episodes and it still got your audience wrong, that is a signal about your site, not just about the model.

Everything is editable. I can go in and correct this if it is wrong. That is the practical difference between a brief and a generated document: a brief is something you argue with and amend, and then it is yours.

The audience section is where you find out quickly how well it understood you. Mine said first, solo founders building something now. Most of them already believe documenting matters, have tried and stopped because of the camera, the edit, or having nothing to say. That is spot on.

Second, makers outside software: shops, studios, bands. Same wish to film the start, no interest in editing or writing threads. Also right, and that one matters because it is the part of the market I keep deciding how loudly to talk about.

Then it gives buying signals. Ready to buy when they have started and abandoned a build-in-public thread or channel, and have already connected tools like GitHub, Circle or Calendar. Signals are the useful bit of an audience section, because they are things you can actually look for in a person rather than a demographic you cannot act on.

When you read this section, do not nod along. Ask whether you could point at five real people who match. If you cannot, correct it before you build a month of work on top of it.

The positioning statement is the sentence the rest of your marketing is cut from. Mine reads: for founders who mean to document their build and never do, Underway is the build-in-public app that reads your day, interviews you out loud, and hands back a finished numbered episode. Unlike recording and editing it yourself in tools like Descript or Riverside, you never edit anything.

Notice the shape. Who it is for, what category you are in, what you do, and then what you are unlike. The unlike clause is the one people skip, and it is the one that makes the statement useful, because it tells you what you are not going to say.

The messaging section gives you hero, problem, solution, proof and call to action. My hero is: in five years, you will be glad you recorded this bit. The problem: everyone means to keep a record of it, nobody does. The solution: it reads your day, asks you out loud, episode is ready minutes later. Those are the same lines that are on the homepage, which is exactly what you want. One message everywhere.

The hero line is the one worth testing. It is also, in my case, the unresolved gap: lead on loss or lead on no editing. The brief surfaces the choice and leaves it with me.

The channel section ranks where your time goes rather than listing everything you could do. Mine says Instagram, start here: post each episode short to Underway Live first, then the same cut with its own words on Underway Rich. YouTube: publish every numbered episode plus the search videos from Videos to film, with the pro thumbnails. Meta and X ads, X, TikTok, and search later.

Search is the slow one. Three hundred and nine impressions, only four clicks at the minute, and nothing on page one. The 59 retitled articles need months. It does take a while to rank for these kinds of things, but we will get there. Having that written down stops you judging a channel on the wrong timescale and killing it in week three.

Who else they look at is the competitor section. The thing with all of these is that they are competitors in a way, but none of them actually do what I do. So it is a bit of an open market, which is sometimes a bad idea, there is usually a reason. At the same time I think now is the time to enter the market. An AI-written competitor list is useful precisely because it names who your buyers compare you to, not who you think you compete with.

A brief that stops at strategy gets read once. Mine ends with your next 30 days: 16 things I can do, and I tick them off as I go.

The plan is organised by week with a theme. Week one is one message everywhere: pick the lead message, rewrite the homepage hero and ad page openers to match, film an episode answering a question from Videos to film, cut three Shorts from it. Week two is win the AI answers: list the searches where AI names a competitor instead of you, film those answers, publish each one on its matching buyer page under the question as a heading.

Week three is anyone making anything, which is the secondary audience made into tasks: film an episode about a shop, studio or band using it, post the Short, reply to people describing the problem, email the trialists before day 15 and ask what is missing. Week four is trials over leads: compare each ad page by card-in trials rather than leads and pause the weakest.

Four weeks is the right horizon for a brief like this. It is long enough to see whether a message lands and short enough that you rewrite the brief before it goes stale. Then you run it again, because by then the AI has more episodes and more shipped work to read.

It suits you if you are a founder or a maker with a live build: a site up, work shipping, some episodes or posts behind you. The brief is written from evidence, so the more trail you have left, the better it is. It also suits people who have been putting off writing a brief for months, because this removes the blank page entirely.

It suits teams who want one shared set of words. The positioning statement, the hero, the problem and solution lines come out in a form you can paste into a homepage, an ad opener and a pitch without rewriting them each time.

It suits you less if you have not built anything yet. With no site, no commits and no episodes to read, the AI has to infer most of it, and you are back to describing yourself in the abstract. Fill in what you can and expect the first brief to be a draft you argue with.

And if you want a long agency-style brand strategy document with workshops and a brand architecture, this is not that. It is a working brief: who, what you say, where you spend, what to do next month.

Hopefully that helps, especially if you are a founder or building a business. It will help you form your marketing message, and it will also show you how clever and intuitive it is at knowing and understanding what you are actually doing.