Stop Buying AI Prompts…

Just Talk to the Bloody Thing

I keep seeing people selling collections of AI prompts, and I have to admit that I find the whole thing rather fascinating. Not because prompts are useless…they aren’t…but because there seems to be an entire little industry growing around the idea that ordinary people somehow need to acquire a secret vocabulary before they can have a useful conversation with a machine.

And we’re not talking about the occasional useful template here. There are entire collections containing hundreds, sometimes thousands, of carefully assembled prompts covering everything from writing and marketing to business, productivity, social media, research and apparently every other conceivable activity that might be performed between breakfast and bedtime. The implication is that somewhere inside this enormous digital filing cabinet lies the magic sentence that will finally persuade artificial intelligence to reveal its true potential.

It is an interesting proposition, particularly when you stop and think about what conversational AI actually is.

The whole point is that you can talk to it.

That sounds almost absurdly obvious, yet somewhere along the way we seem to have convinced ourselves that using AI properly requires us to know the precise combination of words capable of unlocking the machine. The “perfect prompt” has become a sort of digital incantation. Give the machine the correct instructions, arranged in precisely the right order, and enlightenment will apparently emerge from the other end.

There is certainly value in learning how to communicate effectively with AI. Context matters. Clear instructions matter. If you ask for something vague, you may receive something equally vague. If you explain your objective, provide useful information and describe the sort of result you want, the chances of getting something worthwhile improve considerably.

But there is a world of difference between understanding how to communicate effectively with AI and believing that you need a catalogue of pre-written magic spells before you can even begin.

The search for the perfect prompt

Perhaps the attraction is understandable. Technology has always had a habit of making people feel slightly inadequate when they first encounter it. We assume somebody else knows the secret, and if only we could discover what that secret is, we’d finally be doing it properly.

That uncertainty creates a market.

Tell someone that successful AI use depends upon mastering sophisticated prompting techniques and you have created a problem. Sell them a collection of sophisticated prompts and you’ve created the solution.

It is a beautifully simple business model.

And I don’t mean that entirely cynically. There are people who genuinely understand AI systems, have spent considerable time developing effective workflows and can save others an enormous amount of time by sharing what they’ve learned. A good prompt template can be useful, particularly when you’re performing a task repeatedly.

The problem comes when the template becomes the product rather than the expertise behind it.

It reminds me slightly of the old computer magazines that used to publish pages of mysterious commands. You would type some incomprehensible string of characters into the machine, press Enter and hope you hadn’t just instructed it to delete the family photographs.

Back then, knowing the correct command mattered because computers were wonderfully literal creatures. They weren’t particularly interested in what you meant. They were interested in what you told them to do.

Conversational AI changed that relationship.

We can now start with an incomplete thought. We can explain what we’re trying to do, even when we aren’t entirely sure what that is ourselves. We can change direction, correct an assumption, add information, reject an idea and try again.

We can say, “No…that’s not quite what I meant.”

And the machine doesn’t slam the door and go home.

That is rather the point.

The conversation becomes the prompt

This is the bit I think gets missed in the current enthusiasm for prompt libraries.

A conversation with AI doesn’t have to consist of one perfectly engineered instruction followed by one miraculous answer. In many cases, the conversation itself becomes the instruction.

The first message establishes the general territory. The next one narrows it. Another adds context. Another corrects an assumption. You might provide an example of what you mean, followed by an explanation of what you definitely don’t mean. Gradually, the AI has a much clearer understanding of what you’re trying to accomplish than it could possibly have had from your opening sentence.

Humans work like this all the time.

If I sit down with another writer and say, “I’ve got an idea for an article about AI prompts,” I don’t need to arrive carrying a twenty-page briefing document explaining the precise structure of the finished article before we can discuss it. We talk about it. I explain what I’m thinking. They ask questions. I clarify something. They suggest another angle. I reject it. Something else comes up.

Perhaps I realise that I’m not really interested in prompts at all. What interests me is why people feel they need them.

Suddenly the subject has changed.

The conversation has discovered the article.

That is one of the most useful things conversational AI can offer, particularly when you’re doing creative or analytical work. You don’t always know the destination when you begin. Sometimes you only discover it by walking around the subject for a while.

A rigid prompt assumes the destination is already known.

Conversation allows you to find it.

Sometimes the problem isn’t the prompt

This is where things become even more interesting, because sometimes the thing we think we need isn’t actually the thing we need at all.

Someone might believe they need a better prompt for writing social media posts when the real problem is that they haven’t decided what they want their social media presence to achieve. Someone might search for a prompt that will improve their website when the underlying difficulty is that they haven’t worked out what the website is actually supposed to do for the visitor.

Someone might buy a collection of prompts for writing articles when their real problem isn’t writing. They simply haven’t found an interesting question worth writing about.

A prompt can’t necessarily solve an undefined problem.

Conversation can sometimes expose the fact that the problem was never properly defined in the first place.

That distinction matters because AI is increasingly capable of participating in the thinking process rather than merely responding to an instruction. You can ask it to challenge your assumptions. You can ask what information is missing. You can ask it to suggest different ways of looking at a problem. You can tell it that you’re not convinced by its answer and ask it to reconsider.

None of that requires a 2,000-word master prompt.

It requires conversation.

And perhaps that’s why the obsession with perfect prompts feels slightly backwards to me. It treats AI as though it were a machine waiting for the correct command, rather than a conversational tool with which you can gradually work out what you’re trying to accomplish.

The prompt industry isn’t entirely wrong

There is a danger of taking this argument too far.

It would be very easy to write an article declaring that prompt engineering is dead, prompt libraries are worthless and everybody should throw their carefully constructed templates into the nearest metaphorical skip.

I’m not convinced by that.

If you’re performing the same task repeatedly, a good prompt can save a considerable amount of time. If you’re running a business and have a defined process for turning information into product descriptions, reports or marketing material, a reusable instruction set makes perfect sense. If you’re analysing documents according to a particular framework every week, there is no virtue in reinventing the wheel every time.

But notice what makes those prompts valuable.

They contain a process.

They capture experience.

They establish context, constraints, objectives and expected outcomes.

In other words, the value isn’t necessarily in the cleverness of the words. The value is in the thinking that went into constructing the workflow.

That’s an important distinction.

A genuinely useful reusable prompt is rather like a recipe. It tells you what you’re working with, what you’re trying to produce and the method for getting there.

A generic collection of thousands of prompts is rather more like owning a cookbook containing every recipe ever written and still standing in the kitchen wondering what you actually fancy for dinner.

At some point, you still have to open the fridge.

The strange psychology of collecting prompts

I suspect there is another reason these enormous prompt collections appeal to people, and it has very little to do with technology.

Collecting feels like progress.

You download 500 prompts and suddenly you possess 500 possible solutions. Download 5,000 and, presumably, you have become some sort of AI Swiss Army knife.

Except that possessing solutions isn’t the same thing as knowing which problem you’re trying to solve.

We’ve all done versions of this. We buy notebooks because we’re going to become organised. We download productivity apps because tomorrow will obviously be different. We purchase books about writing rather than actually writing. We acquire equipment for hobbies that last approximately three weeks before the equipment ends up in a cupboard quietly judging us.

AI prompts can become another version of the same behaviour.

Acquiring the tools feels productive.

Using them is where things get messy.

And AI is particularly vulnerable to this because the technology still feels unfamiliar to many people. A large collection of prompts creates the comforting impression that somebody else has already worked everything out.

You don’t have to understand the technology.

You just have to find the right prompt.

Except that real life doesn’t usually work that neatly.

The AI guru economy

There is a broader phenomenon here that interests me almost as much as the prompts themselves.

Whenever a new technology arrives, an economy grows around people’s uncertainty about how to use it.

The internet produced web designers, SEO consultants and online marketing experts. Social media produced social media consultants. Cryptocurrency produced people who seemed remarkably certain about the future immediately before the future changed its mind.

AI was always going to produce something similar.

There are genuine experts in this field. There are people who understand the technology deeply, have experimented extensively and have developed workflows that can produce real value. Their knowledge is worth something.

But there is also a difference between selling expertise and selling the appearance of expertise.

A collection of 10,000 prompts looks impressive.

The number itself does much of the marketing.

It doesn’t matter particularly that prompt number 9,847 may be little more than “Write a polite email thanking someone for their assistance.”

Ten thousand sounds better than ten.

And because a digital product doesn’t require a warehouse, a delivery driver or a forklift, there is almost no practical limit to how many of these things can be bundled together.

We’ve therefore ended up with a rather curious situation in which people can spend money acquiring enormous collections of instructions for a technology that can often help them construct the instructions they need themselves.

It is a little like buying a telephone directory containing every possible sentence you might ever want to say to another human being.

Technically impressive.

Slightly missing the point.

Perhaps the real skill is thinking out loud

The more I use conversational AI, the more I wonder whether the skill we should really be teaching isn’t prompt engineering at all.

Perhaps it’s the ability to think out loud.

That sounds considerably less glamorous, which may explain why nobody has yet turned it into a 47-module course with a countdown timer and a limited-time bonus workbook.

But it is useful.

Start with the idea you’ve got. Explain it as best you can. Tell the AI what you’re trying to achieve, even if your explanation is clumsy. Ask what you might be overlooking. Let it suggest possibilities. Challenge the suggestions. Correct it when it gets something wrong. Add information when you realise it matters. Change your mind when a better idea appears.

That is not somehow using AI badly.

It may actually be the most natural way of using conversational AI.

After all, this is broadly how human beings think with one another.

We talk.

We explore.

We disagree.

We refine.

We occasionally discover that the thing we were arguing about at ten o’clock wasn’t actually the thing we cared about at half past ten.

The conversation changes the thinking.

AI can now participate in that process.

Perhaps we should use it.

The prompt may eventually become the conversation

I don’t think prompts are going away, nor should they. Structured instructions will remain important wherever precision, consistency and repeatability matter. Businesses will build sophisticated AI workflows. Developers will continue constructing detailed system instructions. Specialists will develop methods for getting predictable results from particular models.

But for ordinary people trying to write something, understand something, develop an idea, solve a problem or simply make sense of a complicated subject, there is little reason to imagine that every interaction needs to begin with a perfectly engineered block of text.

You can simply start talking.

You can say what you’re thinking.

You can admit that you’re not sure.

You can ask the AI to help you work out what the question should be.

You can disagree with the answer.

You can change direction.

You can start again.

And somewhere in that slightly untidy process, you may arrive at something considerably more useful than the answer you thought you were looking for.

Which leaves me wondering whether the strangest thing about the modern prompt industry isn’t that people are selling thousands of prompts.

It’s that people are buying them while sitting in front of a technology that can often help them work out what they need to ask in the first place.

Perhaps the next time somebody offers you the ultimate collection of AI prompts, containing every instruction you’ll ever need for the rest of your natural life, it might be worth resisting the urge to reach for your wallet.

Open the AI instead.

Tell it what you’re thinking.

If you don’t know what you’re thinking, tell it that.

Ask it to help you figure it out.

You might discover that the perfect prompt was never the point.

The conversation was.

Until Next Time