A founder going blank in ChatGPT, and the brief that replaced it

Going blank in front of ChatGPT is normal for a founder who knows the business cold. One client wrote about exactly that. What changed was a method. Pick one live problem, write a brief that describes it properly, then reuse that brief every week the work comes back.

What can ChatGPT actually do for me?

It does work you can describe. That is the whole boundary, and most of the disappointment sits on the wrong side of it. For a founder with two partners and nobody in marketing or analytics, three kinds of work sit inside the boundary. Research you would otherwise pay an agency for. Drafting you currently do yourself at 11pm. Data you already have and never sit down with.

Launch markets are the clearest case. One brief in my prompt library scores Nutraverde's candidate launch markets on acceptability, price point, customer acquisition cost and regulatory friction. The output is a ranked table with the reasoning behind each score. You read it the way you read a new analyst's first memo. Check two scores in the market you know best. Ask where each number came from. Where the source is thin, that line stops being an answer and becomes a question for a distributor.

The blank screen is a missing brief

A founder goes blank because the question has not been written down anywhere, not because the typing is hard. You carry the problem as a feeling about the week. The tool reads text. Nothing bridges that until you put the problem into words, with the constraints you already hold in your head.

Here is how one client described it. "Before this, I'd open ChatGPT and go blank. The program was fun, non-judgemental, and designed around my actual work. Now I know exactly where and how to use AI."

Read the order in that line. The work came first. The tool came second.

What did the first problem look like?

The first problem is one the founder already owns. I don't know which problem this client picked. Here is what I ask a founder to bring. Something that recurs, weekly or monthly. Something sitting on the desk this week. Something where you already know what a good answer looks like, so you can tell a weak output from a strong one.

That last test matters most. If you cannot grade the answer, you cannot use the tool on that problem yet. Pick the one you can grade. A weekly competitor scan qualifies. A market-entry bet that decides the next two years does not, not on day one.

Two things come out of that first sitting, and neither takes longer than one. A workspace you can work inside, with your own files in it. One problem picked, named, and written down.

A reusable brief is the thing you actually keep

The brief is the asset, not the answer. An answer is worth one week. A brief you wrote once and reuse is worth every week the work returns, and recurring work is most of what a small founding team does.

A brief has five parts: the role the tool is writing in, the context of your business, the task, the format you want back, and the constraints. Constraints are where an operator earns the output. Your categories. Your markets. The claims you are not allowed to make. The sources you will not accept. Write those down once and the output stops being generic on the second run.

Session 2 of Module 1, AI Essentials, is built to end with one of these in your hands, for your own recurring work. That is what A.D.A.P.T. AI Coaching runs on, your live problems rather than a syllabus.

What else do founders ask before they start?

I already use ChatGPT every day. What would change?

The output you keep would change. Daily use without a brief produces answers you rewrite anyway. A brief with your context and constraints turns the same tool into something that returns work you can sign off on, and it returns it the same way every week.

Do I need a paid ChatGPT account?

Yes. A ChatGPT Plus subscription is the prerequisite for working this way. The work runs on your own documents and data inside a workspace you keep, and the brief you reuse every week lives there. Typing a fresh question into an empty box each time is the part you are leaving behind.

The output looks confident and it is wrong. How would I catch that?

You check it before you use it. Take three claims out of the output and trace each one to a source you trust or a person who would know. Where a claim has no source behind it, it goes on the list of things to ask your distributor or your accounts person. Learning to check is the real work.