Have you ever asked AI for something, read what came back, and thought, that is not what I meant?
Did you try once more, get something just as flat, and quietly decide this is not for you?
I would like to give you back those two tries.
The distance between what you asked for and what you actually wanted is a lot shorter than it looks. Closing it does not take a course, a framework, or a tool you do not have yet. It takes one change to how you open the dialog with AI.
Start with the outcome you want.
You gave it a starting point. It needed a destination.
When you type "do my expense report," you have handed over a starting point and nothing else. Which card? What counts as a receipt? How does a client lunch get coded? What happens to the charge with no receipt at all? When is it finished?
It will guess. It will guess generically, because generic is the safest guess available. You will get a neat list of categories and a few questions back, and you will still be the one doing the expense report.
So turn the sentence around and lead with where you want to land.
"I need my October expense report ready to submit. Pull every charge off the corporate card statement attached. Match each one to a receipt in my email or my phone photos. Code them the way I coded September, which is attached too. Mileage goes in at the standard rate from the log. Anything you cannot match to a receipt, list separately at the bottom with the date, the vendor and the amount, and do not guess at it. It is finished when every matched line is coded and the only things left for me are that short list and the submit button."
That took ninety seconds to type instead of ten. It also replaced the two hours you were going to spend on it, and it will replace them again next month and the month after, because the job is written down now. You do not rewrite it. You change the month.
That is a real one. Right now AI does my expense report, pulls my prospecting lists, logs my daily activity into my CRM, and builds the drip campaigns that go out of it. Not one of those started with a clever prompt. Each one started with me writing down what finished looks like.
Last month I wrote that AI gets useful when you hand it jobs instead of questions. This is the part sitting underneath that. A job only becomes a job once somebody says what finished looks like.
Five things to say before you hit enter
The outcome. What is true once this is done.
Who it is for. The actual person who will read it or use it.
The voice. Yours, plain, warm, formal. Say which.
The limits. How long, what format, what to leave out.
The finish line. How you will know it is right.
You will not need all five every time. The first one and the last one do most of the work.
Try it this week
Pick one thing you already have to write, decide or pull together this week. Before you type a single word, finish this sentence out loud:
This is done when ______.
Then put that sentence first and ask for the rest.
Do it three times. By the third one you will stop noticing you are doing it, and you will have a hard time going back.
Where I can help
This is where I start with every client, and it is the smallest part of the work. From there we map what actually runs your week, decide which pieces get handed off, and build the place where all of it lives and gets done. Then I stay in it with you, training your team, adding the next workflow, and changing the build as the business changes.
You can see what that looks like on the AI Strategist page.
If you want to talk it through, book a free Q&A call. Bring one job you are tired of doing yourself. We will write the outcome for it together.
Here's to starting at the finish!
– Dana
