There is a phrase going around that sounds like it means something. Set up AI on your business. Get your AI working properly. AI implementation for small business.
Ask three people what it involves and you will get three answers, and at least one of them is a folder of prompts.
So here is the plain version of what the work actually is. Five things have to be true at the end. Whether you do it yourself or pay someone, it is the same five. If someone offers to do this for you and cannot tell you which of them they are covering, that is worth knowing before you pay.
One: your business is loaded in, not described
Not a paragraph about your business pasted at the top of a chat. Standing information the tool starts every session already holding.
What you sell in the words you actually use. Who it is for and who it is not for. Your real numbers, rates and typical job sizes. How you talk, taken from things you have already written. Your rules, meaning what you always do and never do. What you are trying to do this year.
Skip this and the answers come back generic, which is not the tool being disappointing. It is the space where your business should be sitting empty.
Two: it can reach your real work
Written context is not enough on its own. The tool has to reach the actual material: the spreadsheet, the enquiry emails, the folder of quotes, the documents you already have.
This is where most do-it-yourself attempts stall, because it is fiddly rather than hard. Which folder. Which account. What it is allowed to see and what it is not. What happens when you are away from your desk.
Worth knowing: connecting a mail and calendar account is usually a separate piece of work rather than something included by default. If someone is quoting you, ask whether that connection is in the price.
Three: the jobs that eat your week are built as routines
This is the part people picture as “using AI” and it is actually the smallest part of the work.
A routine is not a prompt you paste. It is an instruction written once, properly, that holds up when you are not watching it. What it produces, in what format, with what rules, and what to do when something is missing.
The ones worth having are unglamorous. A brief that tells you what today actually needs. A weekly plan. A monthly review that happens whether or not you remember. Something that reads the news in your industry and tells you what matters. The admin layer that normally falls on you because there is nobody else to hand it to.
Four: it gets broken on purpose
This is the step nobody advertises and the one that decides whether any of it is safe.
You feed it the bad version of the input. The spreadsheet with the blank cell. The enquiry with no detail in it. The month where two numbers are missing. Then you watch what it does.
If it fills the gap with something plausible, it is not finished, and you have just avoided sending an invented number to a customer. If it flags the gap and tells you what it could not work out, you have something you can rely on.
I came to this through years in QA, where the whole job is assuming a thing is broken until it proves otherwise. It is an unglamorous habit and it is the entire difference between a tool that saves you time and one that quietly hands you a liability.
Five: you know what to do with it on Monday
The setup is worthless if you go back to typing questions into a blank chat.
This part is short. Which jobs to hand over and which to keep. How to ask for something. What to check before you act on an output. What to do when it gets something wrong, which it will.
How much of this should land on you
Here is where I would push back on most of what is offered.
Your part should be small. Answering questions about your own business, and handing over documents you already have. That is it. You should not be assembling a pack, filling in a workbook, or doing preparation before the real work starts.
If setting it up requires a chunk of your week, the thing you were trying to buy was time, and you have just spent it. That is the test I would apply to any offer, including mine.
For what it is worth, this is how I run it: the intake is a short form, or a half-hour call where I do the intake for you and you talk. Then I build. Nothing gets handed back to you to prepare.
What you should not accept
If what you get at the end is a document explaining what you should do next, that is a consulting report. It is not a setup.
Reports are not useless. But a report hands the work back to you, and the reason you were looking for help is that you do not have time to do the work. So you now own a well written PDF and the same problem.
The test: at the end of this, is there something running, or is there something to read?
How to tell if it was done properly
Three questions, and you do not need to be technical to ask them.
What did you load about my business, and can I see it? If the answer is vague, the first part did not happen.
What did you break, and what did it do? If nothing was tested with a bad input, the fourth part did not happen.
What am I meant to do on Monday? If there is no answer, the fifth part did not happen, and the whole thing will be unused within a month.
That last one is the real test. Most failed AI in a small business is not broken. It works fine. It is simply never opened again.
If you would rather have the five done for you than work out which one you are missing, that is what I do, and I would rather you asked me those three questions than took my word for anything. The short intake at quietcontent.com.au is where to start, or book a call and I will do the intake for you.