Your AI Tool Is Only As Good As What You Ask It

We have been thinking about AI backwards.

For the past two years, the conversation has been dominated by tool wars. ChatGPT versus Gemini. Claude versus Copilot. Which subscription is worth paying for. Which model scored highest on some benchmark most of us cannot interpret. We have treated AI like a piece of hardware, as if the right purchase decision will solve everything. But here is the uncomfortable truth that nobody in the industry wants to say out loud: the tool is almost never the problem. The question is.

The single biggest determinant of whether AI helps you or wastes your time is the quality of what you put into it. And most of us, through no fault of our own, were never taught how to do that well.

The Blank Page Problem Has Not Gone Away. It Has Moved.

Think about the last time you sat down to write a difficult email. Maybe it was a performance review for a struggling employee, or a proposal to a client who keeps saying no. You opened the AI tool. You typed something like: "Write me an email about the project update." You got back something generic, something that read like it was written by a committee that had never met you or your client. You either spent twenty minutes editing it into something usable, or you gave up and wrote it yourself.

That is not an AI failure. That is a prompting failure. And it is almost universal.

The blank page problem, the hardest part of any writing or thinking task, used to sit at the start of the work. Now it sits one step earlier. We have to describe what we want before we can get any help wanting it. That sounds simple. It is not. It requires us to be specific about our audience, our goal, our tone, the constraints we are working within, and the context the AI cannot see. Most people skip all of that, get a mediocre result, and conclude that AI is overhyped. Some of it is. But a lot of what feels like AI disappointment is actually a clarity problem in disguise.

A nurse asking an AI to help draft patient discharge instructions will get something completely different depending on whether she types "write discharge instructions for a hip replacement patient" or "write discharge instructions for a 74-year-old woman who lives alone, has mild cognitive decline, and whose daughter will be helping her for the first two weeks. Use simple language, bullet points, and include three warning signs that should prompt an immediate call to the clinic." Both prompts take thirty seconds. The outputs are worlds apart.

Prompting Is a Thinking Skill, Not a Typing Skill

Here is where the conversation needs to shift. Prompting is not about learning magic words or memorising frameworks with acronyms. It is about thinking more carefully before you type. It is about knowing what you actually want, who it is for, what good looks like, and what would make the output wrong. Those are not technical skills. They are professional skills. They are the same skills that make someone a good manager, a good teacher, a good lawyer.

This is why the professionals who tend to get the most out of AI are often not the ones who read the most about it. They are the ones who bring genuine domain expertise and clear thinking to the conversation. A senior HR manager who deeply understands the emotional dynamics of a redundancy process will prompt an AI far more effectively than a junior analyst who has read every prompt engineering guide on the internet. Expertise gives you the vocabulary to be specific. Specificity is everything.

The practical shift we are recommending is this: before you open your AI tool, spend sixty seconds writing down exactly what you need, who the output is for, what format would be most useful, and what a bad version would look like. Treat the AI like a brilliant new colleague who is talented but has zero context about your work, your clients, or your organisation. You would not hand that colleague a vague instruction and expect a masterpiece. Give them context. Give them constraints. Give them a clear definition of done.

The tool is not the revolution. The quality of your thinking, made visible through the questions you ask, is the revolution.

So here is the question worth sitting with this week: If the AI you are using keeps producing results that feel off, generic, or not quite right, what would change if you assumed the problem was the question, not the answer?