Why your first prompt let you down

You typed something like “write me a marketing plan” into an AI chat. A few seconds later, a wall of text landed back. Headings, bullet points, lines like “identify your target audience” and “utilise multiple marketing channels”, the kind of advice that would sit just as comfortably in front of a hairdresser in Truro as it would in front of you. You skimmed it, felt a bit short-changed, and probably haven’t gone back to it since.

That moment is where most business owners quietly write AI off. Understandable, and also the wrong conclusion. The problem was never the tool. It was the one-line request.

Capable, not psychic

Picture the AI as a new member of staff on their first morning. Properly capable: they can write a clean email, summarise a contract, draft a plan, faster than almost anyone you’ve ever hired. What they cannot do, on day one, with nothing to go on, is know a single fact about your business. Not your trade, not your area, not who your customers actually are, not what made your last marketing push work or flop.

So when you hand a new starter one line, “write me a marketing plan”, and walk off, you get exactly what you’d expect from any sharp, willing person with zero information: their best guess at what a generic business in your rough situation might want. Polished. Confident. Not actually about you.

One line is not a brief

Nobody would hand a new employee a single sentence and expect a usable plan back. You’d expect to explain the business first: what you sell, who buys it, what’s worked before, what the budget looks like, when it needs to be ready. Ask AI the same one-line question you’d never ask a person, and you get the same kind of answer a person would give you: something safe, general, and short of what you actually needed.

This isn’t a special AI failing. It’s the ordinary result of an under-briefed request, and it plays out exactly the same way on smaller, everyday jobs. “Write a quote email” gets you a stiff paragraph about “quality workmanship” that could be selling kitchens, sheds or driveways. “Write a job advert” gets you “passionate individual”, “fast-paced environment”, nothing that sounds like your cafe on your street. Same pattern every time: the shorter and vaguer the ask, the more generic the answer, because the AI has nothing else to work with.

The fix is not a better tool

Swapping to a pricier AI model does not fix this. A cheap tool given a proper brief will beat an expensive one given one line, every time, because the deciding factor was never how clever the model is. It’s how much you told it.

That’s genuinely good news. It means getting better answers isn’t about spending more or learning to code. It’s about learning to brief the way you’d brief a real person on their first day, which is exactly what the next lesson is about.