You’ve heard it so many times it’s become background noise: AI is changing everything. Your industry newsletter says it. Half of LinkedIn says it. Someone at a barbecue mentioned their fourteen-year-old is using it for school, effortlessly, and something in your chest tightened a little.
Because underneath all that noise there’s a quieter voice, and it’s yours:
“I really do need to start this AI thing. But I’m totally lost. I don’t even know where to start.”
Maybe you’ve opened one of the tools and stared at the empty box, not knowing what to type. Maybe you’ve got a project sitting on your desk right now — the one you keep thinking surely AI could help with this — and it’s still sitting there, untouched, weeks later. Maybe you asked someone technical and got an answer with so many acronyms in it you quietly decided to work it out later. That was three months ago.
If any of that is you, I want to tell you two things before anything else. You are not behind in the way you fear you are. And what’s stopping you is almost certainly not what you think it is.
I say this as someone from the far side of the gap — most days I have half a dozen AI projects open at once; at this point it’s less a tool I use and more the water I swim in. Yet in the past few weeks I’ve had several conversations with smart, capable people who haven’t typed a single word into an AI tool. Not because they don’t care. One had spent the better part of a month trying to start, without ever getting past the front door.
To make this concrete, let me introduce a professional. He isn’t one person — he’s a persona, a blend of the people behind those conversations — but everything he runs into below came up in a real call. See how much of him you recognise.
Nobody would give him a straight answer about privacy
This professional had something deeply personal to work on. Words that mattered enormously to him, that he didn’t want floating around the internet. So before he pasted them anywhere, he wanted to know one simple thing: is this private?
He did what careful people do. He read the terms and conditions. Then more terms and conditions. Weeks disappeared into legalese, and he came out the other side more anxious than when he went in — having learned a lot, as he put it, but with the job still not done.
Here’s the thing: the straight answer he was looking for exists, and it’s two sentences long. On a paid plan, with “use my data for training” switched off — a single toggle in the settings — your conversations stay yours. That’s it.
Nobody had said that to him in plain English. The industry’s answer to “is it safe?” is a forty-page policy document, which to a normal, careful person reads as no. If you’ve ever closed the tab thinking I’ll figure this out later, that wasn’t laziness. Nobody had handed you the two sentences either.
Everyone in AI seemed to be talking to someone else
He’d asked around before he found me. Everyone he reached was doing AI for corporations — coding teams, accounting firms, legal practices, “massive systems.” When he described what he actually wanted — help seeing the themes in something he’d already written — the response was some version of: Sorry, what? We don’t do that.
Think about how that lands. You arrive with the most human project imaginable, and an entire industry looks at you blankly because it isn’t an enterprise rollout. You start to wonder if the problem is you — if your thing is too small, too odd, too personal to count.
It isn’t. Personal projects are some of the best uses of AI I’ve ever seen. A tribute that has to be compressed to a tenth of its length without losing its heart. A hundred pages of research that needs to become one coherent document. A small business drowning in admin that no software package quite fits. These aren’t edge cases. They’re the point.
He was afraid it would take the pen
This was the deepest one. His conditions were firm: the words had to be his. No clichés, nothing flowery, nothing invented. For the factual documents he works on, he put it perfectly — he couldn’t afford AI creatively deciding that something happened when it didn’t.
He thought these instincts made him unsuited to AI. They’re actually the exact instincts that make someone good at it.
The healthiest way to use these tools is precisely the way he’d already imagined: you keep the pen. AI reads what you’ve written and shows you the threads running through it. It organises, compresses, finds the structure, asks you good questions. It coaches. You decide every word that survives. And anything factual gets checked against the source — always.
So if part of you is wary of AI because you don’t want it flattening your voice or inventing facts — keep that. It’s not a reason to stay off the field. It’s the instinct that will make you good on it.
The gap isn’t ability. It’s translation.
Here’s what I keep coming back to: nothing separating him from me was about intelligence or technical skill. He’s a stronger writer than I am. What separated us was that I’ve had thousands of hours inside these tools to learn what’s safe, what’s possible, and what to ignore — and no one had spent even thirty minutes translating that for him.
The field is moving so fast that the people inside it have stopped noticing the moat forming around it. We answer “can I trust it?” with architecture diagrams. We answer “where do I start?” with a list of seventeen tools. Meanwhile someone with a genuine, urgent, human use case stands at the edge for a month.
If that someone is you: the lostness you’re feeling isn’t a verdict on your ability. It’s a translation failure — and translation failures are fixable in an afternoon.
If you want to start this week
The advice I gave him fits in five lines, and it’s the same advice I’d give you:
- Pick one tool and pay for it. The entry-level paid plan of Claude or ChatGPT is enough — it gets you the good models and the privacy controls. Don’t agonise over which; either will astonish you.
- Turn off “use my data for training” in the settings. One toggle. Now your work stays yours.
- Bring one real task, not a test. Don’t ask it trivia to see if it’s clever. Give it the actual thing on your desk, with the actual constraints.
- Talk to it like a sharp colleague. Explain what you want and why. When the result isn’t right, say what’s wrong and go again. It will meet you more than halfway — these tools are surprisingly good at coaching you into asking better.
- Keep the pen. Let it organise, mirror, and structure. The words, the decisions, and the facts stay yours.
And if you’d like company for the first step
More than one of those calls has ended with me saying: honestly, I don’t think you need me — have a go, you’ll be pleasantly surprised. Sometimes that’s the truth, and I’d rather say it than invoice someone for it. Maybe you’ll read this, have a go tonight, and never need to talk to me at all. Genuinely: good.
But sometimes what you want is an hour with someone who does this all day. Someone to look at the thing on your desk and say “here’s how I’d approach it, here’s what to paste in first, here’s the trap to avoid” — and stay on the phone while you take the first step.
If you’ve got a “surely AI could help with this” sitting in front of you — a personal project, a small business, a mountain of documents, anything — send me what you have, in whatever state it’s in. I’ll tell you honestly whether you need help at all, and if half an hour is all it takes, that’s a good outcome for both of us.
The field’s open. Walking on is easier than it looks.