Free · reads in about six minutes
There's a question in my build quiz that asks how you're feeling about AI right now. One of the answers you can pick is "tried it, it fizzled." I wrote this page for whoever would pick it.
When it fizzles, it's one of five things. Here they are, and what to do about each one.
A free read by Lauren Albert
Enter your email and the read unlocks. A link lands in your inbox so you can find it again, and you'll start getting The Party, my Friday newsletter. Unsubscribe whenever.
The five, at a glance
01
You opened a chat and asked it to write a post, or an email, or a plan. It had never heard of you. It didn't know what you sell, who buys it, or how you talk. So it wrote something that could belong to anyone, and you read it and thought: this isn't me.
That's the correct output for the input. A new employee with no onboarding writes exactly the same email.
The fix
Write eight lines about your business once, and paste them at the top of every new chat before you ask for anything.
Here is my context. Use it for everything in this chat. 1. What I do: 2. Who I do it for: 3. What they pay me for: 4. Three words for how I sound: 5. Three words I never use: 6. What I'm working on this month: 7. What I'm bad at / avoid doing: 8. What a good answer looks like to me:
Keep it in your notes app. Paste it, then ask. The difference is not subtle.
Ten minutes, once.
02
You wrote one caption with it. Then you closed the tab and wrote the next forty by hand. Nothing actually left your plate, so nothing changed, so it didn't stick.
One-off help feels good and disappears. The thing that changes your week is a job that comes back every week.
The fix
Stop asking "what can AI do?" and answer this instead: what did I do this week that I will also do next week, and the week after?
The client questions you answer over and over. The follow-up you keep forgetting. The notes from every call that you never turn into anything. Those are the ones. Boring and repeated beats clever and one-time, every single time.
One sitting. There's a scored version of this at the bottom of the page if you want help picking.
03
You tried it first on the work you're best at. The thing that took fifteen years to get good at, that runs on judgment and taste and knowing the client. It came back shallow. You concluded the tool doesn't work.
You didn't test the tool. You tested it against your strongest skill, and then you graded it as an expert. Of course it lost.
The fix
Point it at the thing you'd be a little embarrassed to admit takes you an hour. Reformatting a spreadsheet. Turning a voice memo into a to-do list. Writing the same confirmation message for the ninetieth time.
Win there first. The judgment work stays yours. That's the arrangement.
Twenty minutes.
04
It gave you something too formal, or too chirpy, or just off. You closed it. That was the whole experiment.
The first output is a draft, not a verdict. Nobody would hire a person, read their first attempt, and fire them without saying what was wrong.
The fix
Say what's wrong in one plain sentence and ask again. Not a better prompt. Just the note you'd give a person.
Too formal. I don't say "delighted" or "reach out." Here's how I actually wrote this last time: [paste] Do it again in that voice.
Three rounds of that gets you further than an hour of prompt engineering. This is the part almost nobody does, and it's the whole difference.
Two extra minutes per task.
05
Something finally worked. A good prompt, a good structure, the exact wording you wanted. It's now buried in a chat thread you'll never find again, so next month you start over from nothing and it feels like AI didn't save you anything.
It didn't. Nothing you build survives in a chat window.
The fix
Make one file. Call it whatever you want. Every time something works, paste it in with two words about when to use it.
WORKS.md - client intake questions → the brief (use after a discovery call) - my voice brief (paste at the top of every new chat) - the follow-up message for people who went quiet
That file is the actual asset. Everything I build for clients is a grown-up version of that file, and the first one always starts in a notes app.
Five minutes, then thirty seconds a time.
Picking the right recurring job is most of the work, and it's the part people get wrong. There's a twelve-task audit that scores yours and names one to start with, in about four minutes. It's already unlocked for you.
Find my one task →The party started. You're not late.