Module 8 of StartCloud's AI at Work: Foundations learning pathway, in five short units with a knowledge check: a practical way to audit your own working week for tasks worth handing over, the five shapes of task AI genuinely handles well (drafting, summarising, reformatting, explaining, brainstorming), the tasks to keep for yourself, and how to build one small habit rather than changing everything at once.
Where AI Actually Fits in Your Job
Audit your own week
Most people meet AI the wrong way round. A tool turns up, everyone has a play, and then the search begins for something useful to do with it. That is how you end up with a very expensive way to write birthday messages.
Start at the other end. Your week already knows where the problem is. For one normal working week, keep a scrappy list on your phone. Every time you catch yourself doing something repetitive, or something that takes far longer than it should, jot down what it was, how long it took and how often it comes around. No categories, no app, no neat handwriting. A running note is plenty.
At the end of the week you will have somewhere between ten and thirty items. Run each one past four questions.
Something you do every Monday is worth ten times more attention than something you do every June. Frequency is what turns a small saving into a real one, and it is also what gives you enough repetitions to get good at asking for the thing you want.
A task that takes an hour of hard thought and ten minutes of typing is not an AI task. A task that takes five minutes of thought and fifty of formatting, rewording and tidying almost certainly is. Split the job in your head before you judge it.
Some work needs your judgement, your relationships or your reputation. That is not a chore, that is the job. Look instead for the packaging around it: the write-up, the summary, the version of the same thing for a different audience.
This is the real gate and almost nobody applies it. If you could spot a mistake in seconds, hand it over cheerfully. If checking would take longer than doing it yourself, or if you simply could not tell, then there is no time saving here, just risk with a nicer font.
Frequent, mostly mechanical, not much riding on it, and obvious the moment it goes wrong. That is where you start, and it is usually the least glamorous item on your list. The exciting ideas can wait until the boring one is running smoothly.
Once you have your list, it helps to know what a good candidate looks like. Across every job we see, the tasks AI genuinely helps with fall into five shapes. Job titles vary wildly. The shapes do not.
Drafting
The blank page is the expensive part, not the writing. A first pass at a client email, a job ad, a policy, a project update, an agenda. You are almost certainly a better editor than you are a starter, and a rough draft to react to gets you moving in seconds.
Summarising
A forty page report you need the guts of before Thursday. The email chain you were cc'd into while you were away. Twelve pages of workshop notes turned into five bullet points and three actions. Long thing in, short thing out, with you checking the short thing against what you already know.
Reformatting
Scribbled notes into a table. A table into a paragraph a client can read. A rambling voice memo into an agenda. The same update written once for the team and once for the board. Nothing new is being created, it is just being poured into a different shaped glass, and that is exactly the sort of work that eats afternoons.
Explaining
A clause in a supplier agreement, in plain English. The error message that says nothing useful. A spreadsheet formula somebody left behind in 2019. Ask for it twice, once simply and once with the detail, and you will understand it well enough to ask a proper question of the person who actually owns it.
Brainstorming
Twenty subject lines so you can pick two. Ten questions to ask a candidate. Five angles on a proposal you have been staring at since Tuesday. You are buying volume here, not quality, and volume is precisely what a machine is happy to produce at eleven at night.
In every one of them, you already know what good looks like. You can read the draft, the summary or the table and tell within seconds whether it is right. That is the actual test, and it is a much better one than the name of the task. If you would not recognise a wrong answer, it does not matter which shape it is. It is not a job to hand over.
A list of what to hand over is only half a plan. The other half is knowing what stays with you, and being able to say why. This is the unit that keeps you out of trouble and, honestly, keeps you good at your job.
None of these are rules handed down from on high. They are just the places where handing work over costs more than it saves, and most of them cost you something you will not notice for months.
The diagnosis, the design call, the read on a client, the decision that needed twenty years of context. Automate the paperwork wrapped around your expertise, not the expertise. If you hand over the core, you have not saved an hour, you have quietly changed what you are selling.
Legal wording, tax positions, safety and compliance detail, technical specifics outside your patch. Not because AI never gets these right, but because you would have no way of knowing when it did not. Unverifiable output is not a time saving, it is a bet.
Module 4 in one line: if you would not email it to a stranger, it does not go into a chatbot. Plenty of otherwise perfect candidates fail on this alone. Sometimes the fix is the right tool on a work sign-in, sometimes the answer is simply no.
Apologies, condolences, thank-yous, hard feedback. Module 5 covered this and it belongs on the keep list. The whole value of those messages is that a person stopped and wrote them, and people can tell.
If you are new to writing quotes, or scoping jobs, or handling a difficult client email, write a few dozen the slow way first. You cannot edit well in a craft you have not learned. The tool will still be there in six months, and you will be far more use with it.
If explaining the task carefully takes longer than doing the task, just do the task. There is no prize for routing a two-line email through a chatbot, and the time you spend perfecting the request is time you did not save.
People sometimes read a list like this as being behind the times. It is the opposite. Knowing precisely where a tool stops being useful is what confident users sound like, in every trade there has ever been. The person who hands over everything is not further ahead than you. They are just further from the wheel.
Here is where most people come unstuck. They finish a course like this one full of intent, try to change eight things at once on the Monday, get a mediocre result on three of them, and quietly stop by Thursday.
The people who end up genuinely quick with these tools almost never did that. They picked one thing.
Go back to your week list and choose the most boring frequent thing on it. The weekly client update. The tidy-up of meeting notes. The same three paragraphs you rewrite for every new enquiry. One task, chosen on purpose, not five chosen by enthusiasm.
Every time, even the day you are busy and it would be quicker to just do it. Two weeks is long enough to get past the fumbling stage and short enough that you will actually finish. This is the whole experiment.
The moment you get a genuinely good result, paste what you asked for into a note. That note becomes your personal library, and it is the bit that compounds. Most people who look effortlessly good at this simply kept their good prompts instead of retyping worse ones from memory.
Did that save time, or did it move the time into fixing the output? Both answers are worth having. If it saved time, you have a permanent habit for the price of a fortnight. If it did not, you have learned something real about your own work, which is more than a demo will ever teach you.
One at a time, forever. A year of that is twenty-six habits, which is far more than anyone gets from a big Monday morning relaunch of how everything works. Slow is not the opposite of ambitious here, it is how it actually sticks.
Tell a workmate what worked. In a small business, the good ideas travel by conversation at the kettle, not by policy document, and one specific example beats an hour of general encouragement. And if your business keeps a list of the AI tools in use, which the National AI Centre recommends, make sure yours is on it. That is not red tape, it is how the business avoids five people quietly solving the same problem five different ways.
And that is the pathway. You know roughly how these tools work, when to trust them, how to ask them better questions, what never goes in, how to own what you send, how to handle recordings and fakes, and now where any of it actually fits in the work you do on a Tuesday. That last one is the piece most training skips, and it is the piece that turns knowledge into a habit worth having.
If you want to keep going, two of our other pathways pick up where this one leaves off.
For teams on Microsoft 365: what Copilot actually is, what the licence buys, and how to roll it out without the mess.
Cyber Hygiene for StaffThe security sibling of this course. Passwords, phishing, updates and the everyday habits that keep a small business safe.
- business.gov.au: Artificial intelligence, a starting point for Australian businesses
- National AI Centre: Guidance for AI Adoption
- Department of Industry, Science and Resources: Voluntary AI Safety Standard
Australia's official guidance on adopting AI safely and sensibly, current at the time of writing (August 2026). All three are written for organisations of every size, small businesses very much included.
Knowledge check
5 quick questions. Get 4 right and the module is yours.
1. What is the sensible way to work out where AI fits your job?
2. Which of these is the shape of task AI genuinely handles well?
3. What is the real test of whether a task is safe to hand over?
4. Which of these belongs on the keep list?
5. What is the best way to actually build the habit?