Welcome: let's start with your week
Here's a quick question. If someone asked what you do for a living, you'd probably give a job title. Teacher, accountant, warehouse supervisor, nurse.
Now a harder one. What did you actually do last Tuesday?
You'd answer that one differently. You'd say things like "answered emails", "ran a team meeting", "checked a spreadsheet for errors" or "calmed down an unhappy customer".
The gap between those two answers is the subject of this whole level. We'll use it all through the course, because it's the most useful way to think about AI and your career.
A job is a bundle of tasks
Let's pin down two words.
A job is the role you're paid for: the title, the contract, the place on the org chart.
A task is one specific piece of work you do as part of that job. Writing a report is a task, and so is booking a delivery or explaining a diagnosis to a patient.
Think of a job as a backpack. From the outside it looks like one thing. Open it up, though, and it's full of separate items. Some are heavy and some are light. Some you use every day and some once a month.
AI doesn't pick up the whole backpack. It reaches in and picks up particular items, a few at a time.
Take a GP practice receptionist. Their week might include answering phones, booking appointments, scanning letters, chasing test results and reassuring a worried patient at the front desk. AI tools could change some of those tasks a lot. Others will hardly change at all. Calling the whole job "automatable" or "safe" misses all of that.
Why researchers count tasks, not jobs
So why does this matter beyond being a neat idea?
Because the way you measure changes what you find. Here's a well-known example.
In 2013, Oxford researchers Carl Benedikt Frey and Michael Osborne scored whole occupations and estimated that about 47% of US employment was at high risk of computerisation. That figure got quoted everywhere.
In 2016, researchers writing for the OECD (Arntz, Gregory and Zierahn) took a task-based approach instead. They looked at what individual workers actually do, since two people with the same job title often spend their days quite differently. On average across 21 countries they found about 9% of jobs were automatable.
Same technology, very different headline, and much of that gap comes from looking at tasks instead of job titles. We'll dig into the studies and why they disagree in later levels. For now, just hold on to this: asking about your tasks tells you far more than asking about your job title.
Augmentation versus automation
Once you're thinking in tasks, you can ask a sharper question about each one. Does AI help you do it, or does it do it instead of you?
Augmentation is when AI helps you do a task. You stay in charge. Say you ask a chatbot such as ChatGPT, Gemini, Claude or Copilot for a first draft of a tricky email. Then you rewrite it, check the facts and decide whether to send it. The AI made you quicker, but the judgement is still yours.
Automation is when AI does the task with little or no input from you. Think of software that transcribes and summarises every meeting on its own, or a system that reads incoming invoices and files them without anyone touching them.
Then there's a third group we shouldn't forget: tasks that are largely unaffected, at least for now. Fixing a leaking pipe, comforting a distressed patient in person, or walking a building site to spot hazards all depend on physical presence, trust or on-the-spot judgement that today's AI tools don't replace.
Here's a subtle point. The same task can sit in different groups depending on how it's done. If you use AI to translate a document and then check it carefully, that's augmentation. If a system translates and publishes it with no human check, that's automation.
There's real-world data on this split. Anthropic, the company that makes Claude, publishes an "Economic Index" that sorts conversations with its tool into augmentation and automation. In its January 2026 report, which used data from November 2025, about 52% of Claude.ai conversations looked like augmentation and 45% like automation. The same report found business use through its API leaned much more towards automation. Remember that this is one company's tool at one moment, and it measures usage, not job losses. Still, it shows that both patterns are happening side by side.
A worked example: Priya, marketing coordinator
Let's try sorting. Priya lists five tasks from her week:
- Drafting social media posts
- Resizing images for different platforms
- Pulling weekly website figures into a report
- Negotiating a sponsorship deal with a local business
- Deciding which campaign ideas to pitch to her manager
How would you sort them? Here's one reasonable answer.
Drafting posts is augmentation. AI can suggest drafts, but Priya knows the brand voice and picks what goes out. Resizing images and pulling routine figures look like automation candidates, since tools can increasingly do them end to end. The negotiation is largely unaffected, because it rests on relationships and reading the room. Choosing which ideas to pitch is mostly unaffected too. AI might offer options, but the call is hers.
Notice what happened there. Priya's job isn't vanishing. Its mix of tasks is shifting.
Your turn: five tasks from your week
This is the exercise we'll build on for the rest of the course.
Grab a note on your phone or a sheet of paper. Think back over a typical week and write down five tasks you actually do. Make each one specific, a verb plus a thing: "reconcile supplier invoices", not "finance stuff". If you're between roles, use your last job or the one you're aiming for.
Next to each task, jot down roughly how much of your week it takes. Then make a first guess: augmentation, automation or largely unaffected?
Don't worry about getting it "right". It's a first draft, and we'll come back to it with better tools in later levels.
Key takeaways
- A job is a bundle of tasks, and AI tends to change tasks one at a time, not whole jobs at once.
- Measuring task by task can give very different results from scoring whole occupations, as the 2013 Frey and Osborne estimate and the 2016 OECD task-based estimate show.
- Augmentation means AI helps you do a task, automation means AI does it instead of you, and some tasks stay largely unaffected.
- The same task can be augmentation or automation depending on how much a human stays in the loop.
- Keep your list of five tasks from your week. You'll need it later.