Start here
Start here with AI
This page is for busy people who keep hearing about AI and want one genuinely useful thing to happen this week, not another tab full of tips. These are the few moves we actually use, in the order we’d walk a friend through them, so save it and share it with anyone who’s been meaning to start.
Three easy first steps
Take them one at a time. A few minutes is enough to start each one, and you don’t need anything you don’t already have.
Stop searching, pick a job
Most people use AI like a search bar, ask a question, get an answer, and move on. Instead, pick one job you do every week and quietly dislike, because that’s where AI starts to give you real time back.
Try this:
I want to stop using you like a search bar and start handing you real work. Ask me short questions, one at a time, about the tasks I repeat every week and which ones I dislike. Then help me choose the single job that would be the best first thing to hand you, and explain why in plain words. Only suggest jobs where you draft and I decide. Nothing that sends, posts or pays on my behalf.
Write a one-page hire brief
Treat it like hiring someone for that one job: write down what the job is, what good looks like, what it must never do, and when it should stop and check with you. If it won’t fit on one page, you’ve probably described two jobs, so cut it back to one.
Try this:
Help me write a one-page hire brief for this job: [describe the job in a sentence]. Interview me one question at a time to cover: what the job is, how often it happens, what a good result looks like, what you must never do, and when you should stop and check with me before going further. Then write the brief in plain words on a single page. If it starts to need a second page, tell me which part is really a separate job.
Run it for a week, then fix the brief
Use the brief on real work each time the job comes up this week, and read everything before you use it. When something comes back wrong, don’t just redo it yourself; add a line to the brief so it gets it right next time.
Try this:
Here is my hire brief: [paste the brief]. Here is this week’s real material for the job: [paste it]. Do the job exactly as the brief describes. Draft only, and don’t send or publish anything. After the draft, list anything you had to guess, and suggest one line I could add to the brief so you won’t have to guess next time.
Your first week
Tick them off as you go. The ticks stay on this device only; nothing is sent.
Where we think this goes
Our prediction: automation becomes a normal part of the human workday.
When Andreessen Horowitz put together its AI Canon reading list, the authors wrote that it’s hard even for experts to keep up with everything new in AI, and “even harder for beginners to know where to start.” That’s the gap this page is for: we think most people don’t need a smarter tool to get going, they need one small, well-described job to hand off first.
Source: Andreessen Horowitz, “AI Canon” by Derrick Harris, Matt Bornstein and Guido Appenzeller · Our own calls. Opinions, not investment advice.
Quick answers
Which AI tool should I start with?
The one you already have. Every prompt on this page is written to work in any of the big chat AI tools, so pick one, stay with it for the week, and judge it by whether the job got easier rather than by which name is on it. You can always switch later, and your hire brief comes with you.
Is it safe to share my data?
Treat it like any outside service you sign up for. Read the tool’s privacy settings before you start, and keep passwords, bank and card details, health information and anything someone shared with you in confidence out of it. A good first job is one where the material is low-stakes anyway, like drafting replies or tidying notes.
How much time does this take?
Each step starts with a few minutes, and the first week is mostly writing the brief and reading what comes back. It feels slower at the start, because you’re teaching the job. The payoff comes later, when the brief is doing the explaining instead of you.
Do I need to know how to code?
No. Everything here is plain writing: a sentence about the job, a one-page brief, and prompts you copy and paste. If you can write a clear note to a new colleague, you can do this.
What if it gets things wrong?
It will, sometimes, and that’s why you read everything before you use it and keep anything that commits you, like money, meetings or public promises, in your own hands. Each mistake is a clue about what the brief left out, so add a line and run it again.
Day One: Don’t chat. Staff.
Most of us start with AI the same way. We open a blank box, explain what we need, get back a tidy paragraph, close the tab, and then explain the whole thing again the next day. It feels useful in the moment, but nothing builds up, and that’s why so many people quietly decide AI is a nicer search bar and not much more.
The shift that changed things for us is simple: stop chatting with it and start staffing it. A chat starts from nothing every time. A staffed job already knows what it’s for, what it should hand you, when it should happen and what it must never do, because you wrote that down once and you hand it over every time.
Here’s what that looks like with an inbox. “Help me with my email” gets you a list of tips you’ll forget by lunch. A staffed version sounds more like this: each morning, sort what came in into act, wait or archive, and draft replies for the ones that need an answer, but never send anything, never archive anything about money or contracts, and never make up a deadline. Same tool, very different result, because the job has been designed.
You still own the parts that matter: taste, risk, relationships and anything that commits you to money, a meeting or a public promise. The AI drafts and you decide. For today, all you need is a notepad. Pick one thing on your desk that repeats and annoys you, and write four lines: the job in one sentence, what should land in front of you when it’s done, when it should happen, and what it must never do. If you can’t name what should land in front of you yet, that’s fine; it just means you’ve found the first thing to figure out.
Follow along on X.
What we're watching in AI and growing companies, as it happens.
Questions? Ask Paul on X: @Paul_Gravette
Paul Gravette · paulgravette.com