Keep your job: use AI to do it better

Learn in your browser · Chapter 01 of 05

The honest, useful version of this question is not about your job title. It is about which of your actual tasks AI can help with, and which stay yours. This chapter sets up the method the rest of the workbook walks you through.

Search for 'will AI take my job' and you will find two kinds of answer: blanket reassurance that says not to worry, or blanket alarm that says everyone is at risk. Neither is very useful, because both talk about jobs as one solid thing, when a job is actually a bundle of different tasks, done in different ways, at different times in the week.

The more useful question is: which of the tasks that make up your week can AI genuinely help with, and which ones need a human doing them, specifically you, with your judgement, your accountability, and your relationships with the people you work with or for? Some tasks in almost every job are AI-assistable. Some are not, and will not be for a long time, whatever a headline promises.

This workbook gives you a method, not a pep talk. Over the next four chapters you will map your own weekly tasks, sort them honestly into what AI can help with and what stays yours, pick two or three workflows to actually try this week, and set up a simple way to measure whether they are worth keeping. It works whether you do admin and office work, a trade, a retail or customer-facing job, teaching, or healthcare support, because the method stays the same even though the tasks are completely different.

• Q1: Is it true that AI is coming for most jobs?
• A: No, not in the way that headline suggests. Research on this, including the World Economic Forum's future of jobs work, points to AI reshaping the mix of tasks within roles and creating new roles alongside the ones it changes, more than deleting whole jobs outright. The risk is greatest for people who ignore it, not for people who learn to direct it.
• Q2: Why does this workbook talk about tasks instead of jobs?
• A: Because a job is not one thing. It is a bundle of different tasks, and AI's usefulness varies hugely between them. Sorting at the task level is the only way to get a genuinely useful, personal answer instead of a generic one.
• Q3: Do I need to already be good at using AI tools to get through this workbook?
• A: No. The method matters more than tool skill. You will pick simple, free AI tools and apply the method to your own real tasks, starting this week.
• Q4: What if I work in a job that is not covered by the examples in this workbook?
• A: The method still applies. The worked examples across admin, trades, retail, teaching and healthcare support exist to show the method in different settings, not to limit it. Map your own tasks in the next chapter and the same method works.

Learn in your browser · Chapter 02 of 05

Generic 'AI skills' advice fails here first: it never asks what you actually do all week. This chapter fixes that with a real, honest task list.

Most people, asked what they do at work, describe their job title or a handful of headline responsibilities. That is not detailed enough to work out where AI helps. You need the actual list: the specific, recurring things you do, roughly how long each one takes, and how often.

Here is a worked example from an office administrator, to show the level of detail this needs.

Scroll sideways to see every column.

Notice the list mixes clearly repetitive, wordy tasks (drafting letters, data entry) with tasks that depend entirely on being a person another person is dealing with (greeting visitors, handling a difficult caller). That mix is normal, and it is exactly what the next chapter sorts through properly. For now, the job is just to get an honest list down.

• Q1: How detailed does my task list actually need to be?
• A: Specific enough that you could hand the description to someone else and they would know exactly what you meant. 'Admin' is too vague. 'Answering routine email enquiries' is specific enough.
• Q2: What if my week is different every week?
• A: Keep the running note for a longer stretch, a week or two, and list the tasks that show up repeatedly even if the exact mix changes. The point is to capture your real recurring work, not one perfect week.
• Q3: Should I include tasks I do rarely, like once a quarter?
• A: Only if they take real time when they happen. A quarterly task that eats a whole day is worth including. A five-minute task you do twice a year is not worth the space on this list.
• Q4: Why bother marking which tasks take the most time?
• A: Because the next two chapters work best when you focus your effort on the tasks that matter most to your week, not on every minor task equally. Time spent is the first filter.

Learn in your browser · Chapter 03 of 05

Not everything on your list should go anywhere near an AI tool. This chapter gives you a clear way to tell the difference, with worked examples across several kinds of job.

AI is genuinely good at some kinds of task and genuinely unsuited to others, and the difference is not about how 'high skilled' a task looks from the outside. It is about what the task actually requires.

Scroll sideways to see every column.

A simple test: if the task is mostly about producing or reorganising wording, structure or a first pass from information you already have, it is usually AI-assistable. If the task is mostly about deciding, being responsible for a real person's outcome, or doing something physical, it usually stays yours.

Worked examples across different jobs. The table below shows one AI-assistable task and one judgement-only task from each of five different job families, to make the pattern concrete rather than abstract.

Scroll sideways to see every column.

• Q1: What is the quickest way to tell if a task is AI-assistable?
• A: Ask what the task is actually made of. If it is mostly wording, structuring or summarising information you already have, it is usually AI-assistable. If it is mostly deciding, being responsible for someone's outcome, or physical skill, it stays yours.
• Q2: What does 'mixed' mean, and why does it matter?
• A: Many real tasks have an AI-assistable part, like the wording, and a judgement-only part, like the decision behind it. Splitting the two means you get the real benefit from AI on the wording without quietly handing over the decision as well.
• Q3: Why is patient or client data called out specifically?
• A: Because it is the clearest, most common way people accidentally misuse AI tools at work: pasting someone else's private details into a public AI chat tool. The task might genuinely be AI-assistable in principle, but not with real personal data attached, ever.
• Q4: I work in a job family that is not in the worked-examples table. What do I do?
• A: Use the general test in the first table, not just the specific examples. The five job families are there to show the pattern working in different settings, not to be an exhaustive list. The same wording-versus-judgement test applies to your role too.

Learn in your browser · Chapter 04 of 05

The point is not to change everything at once. A small number of real, working habits beats a long list of things you tried once.

You now have a sorted task list. The next step is choosing, deliberately, two or three tasks to actually build an AI-assisted habit around this week, not experimenting with everything at once and losing track of what actually helped.

The best candidates score well on three things: they take up real time across your week, they are genuinely AI-assistable or have a clearly separable AI-assistable part, and getting the AI's first attempt wrong would cost you a few minutes of correction, not a real consequence for anyone. Avoid starting with a task just because using AI on it sounds impressive; start with the task that will actually save you time this week.

Worked examples of a first workflow, by job family.

Scroll sideways to see every column.

• Q1: Why only two or three workflows, not more?
• A: Because the point is to build a real, checked habit, not to try a long list of things once and forget most of them. Two or three you actually keep beats ten you tried and dropped.
• Q2: How do I choose which tasks to start with if several qualify?
• A: Favour the ones that take the most time across your week and carry the least risk if the first AI draft needs fixing. Time saved and low risk together make the strongest starting point.
• Q3: What if the AI's first attempt at a workflow is genuinely bad?
• A: That is useful information, not a failure. Either the prompt needs more of your real facts and context, or that particular task is less AI-assistable than it looked. Adjust the prompt once or twice; if it still is not working, that task probably belongs back on the judgement-only side.
• Q4: Do I need a paid AI tool to do this?
• A: No. Free AI chat tools are enough for every workflow in this workbook. The method is what makes the difference, not which tool you use.

Learn in your browser · Chapter 05 of 05

A workflow you have not measured is a guess dressed up as a habit. This chapter gives you a simple way to know if it is actually working, then closes the workbook.

Without measuring, you cannot tell a workflow that genuinely saves you time from one that just feels modern. The measurement does not need to be sophisticated: a rough log, kept honestly for two weeks, is enough to see a real pattern.

Scroll sideways to see every column.

Log the task, your rough time before you used AI (your honest memory or your task list from chapter two), your rough time with AI including any editing or checking, and one short note on quality. Two weeks is usually enough to see whether the saving is real and whether the quality is holding up.

• Q1: Do I need special software to measure the time I save?
• A: No. A simple table on paper, in a notes app or in a spreadsheet is enough. The habit of logging honestly matters far more than the tool you log it in.
• Q2: What if a workflow saves time but the quality drops?
• A: Time saved is not worth it if you are quietly shipping worse work. Treat a quality problem as seriously as a time problem, and fix the prompt or the check before deciding the workflow is worth keeping.
• Q3: How long should I keep measuring before I decide?
• A: Two weeks is usually enough to see a real pattern rather than one unusually good or bad day. Set the review date in advance, as the exercise above asks, so you actually look at the results rather than just assuming it is working.
• Q4: What do I do with a task that turns out not to be AI-assistable after all, once I have tried it?
• A: Move it back to judgement-only on your sorted list from chapter three, and pick a different task from your list to try instead. That is a normal, useful outcome of the method, not a failure of it.

What you can do now. You have a real, honest map of your own weekly tasks, a clear way to sort what AI can genuinely help with from what has to stay yours, two or three AI-assisted workflows you have actually tried on real work, and a simple log that tells you whether they are worth keeping. That is a genuinely different position to be in than either ignoring AI or trying to use it for everything at once.

Keep this workbook. It is yours to keep, print and write in. Come back to chapters two and three whenever your role changes, a new tool appears, or you take on a new set of tasks, and run the method again.

https://trust-agent.ai/exam/keep-your-job-use-ai-better

This is free educational content published by Mickai LTD, not professional, legal, financial or compliance advice, and Mickai accepts no liability for how it's used.

Trust Agent publishes workbooks like this one for free because Mickai believes people who learn to direct AI at their own real work, deliberately and with proper checks, are the ones who stay in control of their role, not the ones AI quietly replaces.