Turn your team's private ChatGPT habits into one proper AI job
Staff already using ChatGPT on their own? How to find the one habit worth keeping and make it a written job that doesn't leave with the person who found it.
I believe your best first AI job may already exist, in one person’s private ChatGPT account. It’s your business, and you never asked the team to use ChatGPT or Copilot, but some of them do, on their own accounts, because it helped.
One person in your team pastes a customer’s request and the price list into ChatGPT and gets back a quote email the customer can act on. Nobody else knows how she does it.
Leave it there and that habit leaves you exposed in two ways. The method vanishes when she’s off or leaves, and customer detail sits in an account the business doesn’t control. I’ll walk you through turning one private habit into a written job the business owns, tested properly and with a named owner.
Ask each person what they already use it for
Start with short, relaxed conversations with the people who use ChatGPT most. In most teams that’s likely to be whoever prices and sends quotes and whoever answers customer enquiries. You want the habit, not a confession, so say at the start that nobody is in trouble.
Ask each person three questions: what do you use it for, what do you paste in, and what do you do with the answer? The first answer shows you the work, the second shows you the risk, and the third shows whether the output reaches a customer, a record or nowhere at all.
Listen for use that repeats around the same piece of work. One person turns a customer’s request and the price list into a quote email. Another drafts the first reply to a website enquiry. A third summarises a month of orders before a customer review. The signal isn’t that someone opened ChatGPT. It’s that they keep coming back with the same kind of input because the answer helps them finish a task you’d recognise.
Write each habit on one line: what starts it, what goes in, what comes out and where it goes next. By the end you’ll have a short list of real uses, found by the people who do the work rather than by a vendor demonstration.
Pick the habit that repeats and can be checked
The habit worth turning into a job repeats often, starts from a recognisable input, produces something a person can check, and matters enough to be worth the effort. It should already help with real work. A habit that exists because someone enjoys playing with the tool isn’t the one.
Quote emails pass that test. The trigger is a quote request arriving, and the input is the request and the price list. The output is a draft someone can check against the price list before the customer sees it. “Helping with customers” fails, because it hides several different jobs with different inputs and different risks.
Do a rough sum for each habit on your list. Say four people each write six quote emails a day and each email takes 20 minutes to write. If those figures were yours, that would be 8 hours of staff time a day on quote emails, before anyone checks them. The sum tells you which habit is worth formalising first.
Prefer work where a person reads the result before it reaches a customer or changes a record. Then choose one. Several habits may look promising, and it’s tempting to write them all up at once. One job done properly teaches you what the brief, the check and the ownership need to look like, and the next job gets easier because of it.
Write down the six parts of the job
A private habit works because one person answers six questions in their head every time. Each time she words it slightly differently and fixes the answer without writing down why, so there’s no version anyone can improve. To make it a job, you write the answers down.
The trigger is the event that starts the work, such as a quote request arriving from a customer. It replaces the need for someone to remember to open ChatGPT.
The input is the information the job is allowed to use. For quote emails that means the customer’s request, the current price list and your terms and conditions, never passwords, bank details or anything copied from the customer’s own systems. Name where the input comes from and which version is current. If something is missing, the job goes back to a person.
The brief is the written instruction: what to produce, in what tone, what to leave out and what to do when unsure. How to brief an AI tool covers this in detail. Ask the person who found the habit to help write it, because her judgement is the part that usually goes missing. She knows which background to include, which terms need explaining and which answer sounds right but isn’t. Writing down the unwritten rules shows how to get that onto paper.
The standard is what the reviewer must find before the email goes out: the right prices and quantities, no internal jargon, and the correct next step for the customer. Agreeing what good looks like before AI goes live shows how the people who do and check the work set that standard together.
The owner keeps the brief and the inputs current, takes the exceptions and can stop the job if it slips below the standard. For quote emails that’s often the sales lead. Who owns the job sets out the authority and time that person needs.
The visible result is where the finished work lands, such as a draft in the shared inbox rather than a private chat window. If the output stays in someone’s browser, the business still doesn’t own the job.
Move the work onto an account the business controls
This step matters because your customers trust you with their details. Quote requests can hold names, addresses, agreed discounts and the odd password someone typed in by mistake. That detail shouldn’t sit in a personal ChatGPT account the business can’t see or close.
Decide which tool and which account the job will run on, then check that account’s terms for how it stores and uses what you send. If you can’t say what happens to the information, the job doesn’t run on customer data yet. How to write an AI use policy sets out what staff may paste in, and the security question gives you the questions to put to the supplier.
Don’t turn this into a crackdown. The people who found the habit did you a favour by showing where the time goes. Move the work somewhere safe, strip out what shouldn’t be sent, and keep them involved in what happens next.
Test the written job with a different person
The job is ready to test when a different person can follow the written version without asking the one who found it. They should know when to start, what to use, what a good email contains, who checks it and where it goes.
Walk through it together first. The original person’s role is to spot the judgement that fell out when the habit was written down. If she has to add a rule out loud, it goes into the brief before the test starts.
Then run it on real quote requests as they arrive. Keep the old way running beside it, and keep a person reading every email before it reaches a customer. Record every correction, because a correction that repeats tells you where the brief, the input or the boundary needs work. How to test an AI job with the people who do it sets out how to run that trial and read the results.
Decide whether the job goes live, changes or stops
At the end of the trial, the decision is about this one job, not about ChatGPT in general. It goes live when ordinary emails meet the standard and the checking time still leaves a real saving against writing them by hand. If it goes live, decide where the saved staff time goes, and what to do with the hours AI saves covers that. It changes when the useful part is clear but the brief or the inputs need work. It stops when the result doesn’t justify the effort, or the approved account can’t handle the information safely.
Whichever way it goes, tell the person who found the habit what happened and why. If the job goes live, she’s a natural choice to own it or to train the others. People keep showing you useful habits when they can see what came of the last one.
A method that once lived in one browser tab can now survive a holiday, a new starter or a cleared history. The business owns it, a named person answers for it, and your customers get emails to the same standard whoever picked up the request.
Book three conversations with your team this week
On Monday, pick the three people you think use ChatGPT most, and bring in your office manager or sales lead if you have one. Book 20 minutes with each and ask the three questions: what do you use it for, what do you paste in, and what do you do with the answer?
Write each habit on one line. By Friday, choose the one that repeats, can be checked, and comes out best on the rough sum. That’s about 90 minutes of your week, and it gives you your first job and the person best placed to help you write it down.

