Where to start

On why the most asked question about AI is the wrong question

There is this moment I've lived through dozens of times now, usually over coffee or in Slack DMs, where someone leans in and asks with genuine urgency: "I know I need to use AI, but where should I start?"

Every founder, marketer, and knowledge worker I know is asking this question. They've read the articles, watched the demos, seen the productivity porn on Twitter. But their first attempt disappointed them. They typed a prompt. Got an answer that was wrong, or generic, or buried in jargon they didn't ask for. Then they closed the tab.

Even people who've spent weeks researching which AI tool to use find themselves back at the same question three months later. Every attempt starts from zero, no matter how much preparation.

The old playbook of "research thoroughly, then commit" doesn't work anymore. AI tools multiplied faster than anyone could evaluate them. The starting line keeps moving, and by the time you've chosen, you're already behind.

Enter the trial-and-error phase. Over the past year, I've watched people attempt to get started in three predictable ways.

Not to integrate AI into their workflow. Simply to feel like they're not falling behind.

Some sign up for ChatGPT, ask it three questions, get three mediocre answers, and ghost it. Others watch YouTube tutorials on prompt engineering, taking notes on frameworks they'll never use. The most ambitious spend weekends comparing features across a dozen tools, building elaborate decision matrices that end in paralysis.

There are people spending hours crafting the perfect prompt template. Not because it produces better outputs. Simply because optimization feels like progress. It's spreading beyond early adopters: executives asking their teams for "AI strategies," consultants selling "transformation roadmaps," everyone performing the theater of getting started.

This shadow economy exists because no one wants to admit they're lost. Everyone invented their own explanation for why AI hasn't clicked yet. If someone else figures it out first, if they crack some secret starting formula, that's an existential threat to feeling competent.

Three months ago, I was convinced this was an adoption curve problem. Give it time. People will figure out their use cases. The tools will get better.

Then I started asking people what stopped them, and the entire pattern changed.

The pattern has three parts, and they all point to the same thing.

First: You expected to ask a question and get the answer. Instead, you got an answer. Close enough to be plausible. Wrong enough to be useless. The gap between demo magic and your reality was wider than you expected.

When you try again with a better prompt, you get a better answer. But it's still generic. Lacks your context. Misses the nuance. You wanted transformation. You got autocomplete.

That's when it hits: The tools work as advertised. You didn't advertise the right problem to solve.

Why spend hours researching the perfect starting point when the starting point doesn't matter? No amount of preparation. No perfect tool choice. Start wrong, then fix it.

I watched a founder spend two weeks evaluating AI writing tools. Notion AI, Jasper, Claude, ChatGPT Plus, Gemini. Feature matrices. Pricing tiers. Integration capabilities. Then he picked ChatGPT, used it twice, and gave up because "it didn't understand my business."

That's the difference between researching AI adoption and adopting AI. Research assumes there's a right answer to find. Adoption assumes you'll find the answer by starting anywhere.

Every tool becomes the wrong tool until you know what problem you're solving.

Second: You're trying to make a permanent decision with temporary information. You see 50 AI tools. Each promising to revolutionize your workflow. Each with technical barriers to entry. Each requiring you to learn new interfaces, new prompts, new mental models.

You want to choose once. Commit. Build your process around it. Switching costs feel enormous. Upload all your documents, train it on your style, integrate it with your tools, then discover it's the wrong fit? You lose everything. The context. The history. The investment.

So you freeze. Waiting for clarity that only comes from the thing you're avoiding: using one.

Third: You want AI to slot into your current workflow. But your current workflow was designed for a world without AI. You're asking "where does this fit?" when you should be asking "what becomes possible now?"

Both the disappointment and the overwhelm point to the same shift: The right tool doesn't exist yet. Not until you've used the wrong one enough times to know what "right" means.

The old approach to technology adoption was about commitment. You evaluated options, made the best choice, trained yourself on the tool, and defended that choice. That model is dying. The new approach is about promiscuity. You try everything. Break things. Switch without guilt. Value emerges from iteration, not optimization.

AI tools are designed to make you feel like you're already behind. Every demo shows someone using it perfectly. Every tutorial assumes you know what you're trying to accomplish. Every "getting started" guide skips the part where you don't know what questions to ask.

They give you infinite flexibility to solve problems you don't yet know how to name. "Set your custom instructions." "Train it on your documents." "Integrate it with your stack."

This is the future nobody's talking about. Not AI as tool, but AI as mirror. It shows you where your thinking is lazy. Where your questions are imprecise. Where you assumed someone else had the answer.

The people who "get" AI aren't smarter. They're more comfortable looking stupid in front of a chatbot.

Every person asking "where should I start" is operating from the same assumption. That there's a starting line. That others found it. That falling behind means you missed something obvious.

We're entering an era where the greatest barrier to using AI lives in your head, not your infrastructure. The willingness to try something that might not work. To look incompetent. To iterate in public (even if "public" is you and a chatbot). To accept that the best use case you discover might happen accidentally while solving something else.

The perfect starting point doesn't exist. Never has.

We're about to watch millions of people talk themselves out of using the most powerful technology shift in decades. Not because it's too hard. Because they're waiting for permission to start wrong.

This is a choice for people building companies, creating content, making decisions. If you're trying to stay competitive, you're already in the game whether you admit it or not. For you, waiting for the perfect starting point is choosing to fall behind.

For everyone else? Your friend who's not interested, your colleague who's happy with their current tools, the person who opted out entirely? They're not wrong for sitting this out. They're spectators, not participants. Most people should be spectators.

But some of you are stuck in between. You want the leverage AI promises but fear looking incompetent while learning it. You're preparing to get started instead of starting. Those are the ones who'll lose. In technology adoption, you're either learning in motion or learning in theory. There's no middle ground.

The few who get it will pick any tool. Use it badly. Switch when something better appears. And watch as their "behind" becomes everyone else's "ahead."

The rest will still be researching starting points while others are already at the finish line.

The answer will be simple. You started.

And then you iterated.

Your confusion? That's the process.

The only question is whether you'll start before you're ready.