If you typed will AI take my job into a search bar late at night, here’s the straight answer. Probably not your whole job – but AI will likely take over some of your tasks, and that’s where the risk sits.
Research keeps landing in the same place: AI automates tasks, not whole jobs. The most exposed jobs are the ones where most of the week is text, data or routine digital work that someone can check quickly. Jobs that depend on physical presence, licensed accountability or trust are the least exposed.
The test below takes 15 minutes. List your weekly tasks, score each one, and see whether your job is likely to be augmented, reshaped or at real risk. Then follow the plan for your result.
No doom. No pep talk. Just a pen, last week’s calendar and an honest look at how you actually spend your hours.
Key terms used in this guide: AI literacy and large language model.
The 15-minute test
Most articles on this topic hand you a list of “risky” and “safe” job titles. The trouble is, nobody does a job title all day. You do a pile of tasks, and some of them are far easier for a model to do than others. So we score the pile.
Step 1: List your 10 most time-consuming weekly tasks (3 minutes)
Be specific. “Emails” is not a task. “Answering supplier questions about delivery dates” is. Open last week’s calendar and your sent folder while you do this – memory tends to flatter us, and the boring stuff is exactly what we forget.
Step 2: Score each task with four questions (8 minutes)
Give every task a 0, 1 or 2 on each question:
- Is it mostly text, data or code? 0 = no, it’s hands, voice or judgment in the moment. 1 = partly. 2 = yes, the input and the output are words, numbers or code on a screen.
- Can someone check the output quickly? 0 = checking takes deep expertise or weeks of waiting. 1 = with some effort. 2 = a colleague can tell in minutes whether it’s right.
- Does it need you physically present? This one is reverse-scored. 0 = always, you have to be in the room or on site. 1 = sometimes. 2 = never.
- Does it need your license, trust or accountability? Also reversed. 0 = yes, your signature, license or relationship is the whole point. 1 = somewhat. 2 = no, anyone could sign it off.
Why these four? The first two describe what makes a task easy to hand to a model: models work in text, and fast checking makes their output cheap to trust. The last two are the brakes. A model can’t crawl into an attic, and it can’t lose its license.
Here’s the worksheet. Copy it into a doc or spreadsheet and fill it in.
| Task | Mostly text, data or code? (0–2) | Can someone check the output quickly? (0–2) | Needs you physically present? (reverse score) | Needs your license, trust or accountability? (reverse score) | Task score |
|---|---|---|---|---|---|
| Example: weekly sales summary for my manager | 2 | 2 | 2 | 1 | 7 |
| 1. | |||||
| 2. | |||||
| 3. | |||||
| 4. | |||||
| 5. | |||||
| 6. | |||||
| 7. | |||||
| 8. | |||||
| 9. | |||||
| 10. | |||||
| Total | /80 |
Step 3: Total the scores (2 minutes)
Each task scores 0 to 8, so ten tasks give you a total between 0 and 80. One tweak worth making: if a single task eats half your week, count it twice and drop your smallest one. Hours matter more than line items.
Step 4: Read your tier (2 minutes)
Now find your total in the table below. Be honest with yourself here – a flattering score helps nobody, and a scary one isn’t a verdict.
What your score means: three tiers
So, will AI replace my job? Your number doesn’t give a yes or a no. It tells you which conversation you should be having – with yourself and, sooner or later, with your manager.
| Score range | What it means | First three actions |
|---|---|---|
| 0–35: Augmented | AI makes you faster. Your core work needs hands, presence or a license, and AI mostly touches the paperwork around it. | Pick your most annoying admin task and try an AI tool on it this week. Learn one tool properly instead of five badly. Track the hours you save. |
| 36–55: Reshaped | Your role changes. A real chunk of your week can be drafted or processed by AI, but the job still needs your judgment and relationships. | Mark which of your tasks scored 6+. Get good at doing those with AI, not against it. Move time toward the low-scoring, judgment-heavy tasks. |
| 56–80: At risk | The core tasks are automatable. Most of your week is text or data work that’s easy to check, with little need for your presence or signature. | Stop waiting for a clear signal from above. Build AI skills that let you run the automated version of your own work. Start mapping an adjacent role with lower scores. |
Augmented (0–35): AI makes you faster
This is where most hands-on, licensed and people-facing jobs land. The main risk isn’t replacement. It’s being the one person on the team still doing reports by hand while everyone else finishes by lunch.
Next step: automate one piece of paperwork this week and time the difference.
Reshaped (36–55): your role changes
This is the messy middle, and plenty of office jobs live here. Your title may survive while the job underneath it shifts – fewer first drafts, more reviewing, more deciding. People who lean into that shift usually end up with the more interesting half of the work.
Next step: take your highest-scoring task and write down how you’d hand it to an AI tool, step by step. That’s your future workflow.
At risk (56–80): the core tasks are automatable
A high score isn’t a pink slip. It means the tasks that fill your week are the easiest kind to automate, so your employer is likely to try. You have more control over what happens next than it feels like right now – but you’ll want to use it early.
Next step: this month, learn to do your own highest-scoring task with AI, then talk to your manager about owning that new process.
Worked examples: the test applied to 5 roles
Numbers on a table are easier to believe when you see them used. Below, we ran the test on five common jobs. The tasks come from the U.S. O*NET task lists for each occupation; we picked four representative ones per role, averaged the task scores and multiplied by ten to estimate a full-week total.
These are illustrative, not official ratings. Your own version of the job may look very different – which is exactly why the test is about your week, not your title. If you’ve been asking will AI take over my job about one of these roles, use the example as a starting point and then redo it with your own tasks.
Customer service representative
| Task | Text/data | Quick check | Presence (rev.) | Trust (rev.) | Score |
|---|---|---|---|---|---|
| Answer routine questions by chat or email | 2 | 2 | 2 | 2 | 8 |
| Log customer interactions and case notes | 2 | 2 | 2 | 2 | 8 |
| Process refunds and account changes | 2 | 2 | 2 | 1 | 7 |
| Calm and resolve escalated complaints | 1 | 1 | 2 | 0 | 4 |
Estimated total: about 68 – at risk, at least for roles that are mostly scripted chat and email. The escalations are the part that holds up, because an angry customer wants a human who can bend a rule. More in will AI replace customer service.
Accountant
| Task | Text/data | Quick check | Presence (rev.) | Trust (rev.) | Score |
|---|---|---|---|---|---|
| Reconcile accounts and flag discrepancies | 2 | 2 | 2 | 1 | 7 |
| Prepare tax returns | 2 | 1 | 2 | 0 | 5 |
| Review and sign off on financial statements | 2 | 1 | 2 | 0 | 5 |
| Advise clients on tax and financial decisions | 1 | 0 | 2 | 0 | 3 |
Estimated total: about 50 – reshaped. Almost everything is data, which pushes the score up. But the signature, the license and the client relationship pull it right back down. The job drifts from producing numbers to standing behind them. See will AI replace accountants.
Marketing coordinator
| Task | Text/data | Quick check | Presence (rev.) | Trust (rev.) | Score |
|---|---|---|---|---|---|
| Draft social posts and email copy | 2 | 2 | 2 | 2 | 8 |
| Pull campaign performance reports | 2 | 1 | 2 | 2 | 7 |
| Track budgets and invoices | 2 | 2 | 2 | 1 | 7 |
| Coordinate vendors, shoots and events | 1 | 0 | 1 | 1 | 3 |
Estimated total: about 63 – the low end of at risk. That surprised us a little, until you look at the tasks: drafts and reports are precisely what current tools do well. The coordinator who becomes the team’s AI workflow person tends to be in a much stronger spot. Read will AI replace marketing jobs.
Electrician
| Task | Text/data | Quick check | Presence (rev.) | Trust (rev.) | Score |
|---|---|---|---|---|---|
| Install wiring, fixtures and panels | 0 | 0 | 0 | 0 | 0 |
| Diagnose faults on site | 1 | 0 | 0 | 0 | 1 |
| Read blueprints and plan the job | 1 | 1 | 1 | 1 | 4 |
| Write quotes, invoices and paperwork | 2 | 2 | 2 | 1 | 7 |
Estimated total: about 30 – augmented. The paperwork is highly exposed; the work itself isn’t. No model is pulling cable through a 1970s ceiling any time soon. Full breakdown: will AI replace electricians.
Registered nurse
| Task | Text/data | Quick check | Presence (rev.) | Trust (rev.) | Score |
|---|---|---|---|---|---|
| Administer medications and monitor patients | 0 | 0 | 0 | 0 | 0 |
| Assess patient condition | 1 | 0 | 0 | 0 | 1 |
| Chart and document care | 2 | 1 | 2 | 1 | 6 |
| Educate patients and families | 1 | 1 | 1 | 0 | 3 |
Estimated total: about 25 – augmented. The one task that scores high is documentation, which many nurses would happily hand off. More in will AI replace nurses.
Notice the pattern? The two most exposed examples aren’t the “least skilled.” They’re the ones where the work lives on a screen.
What the research actually says
The test above isn’t a party trick. It’s a simplified version of how economists study the question will my job be replaced by AI – by breaking jobs into tasks first.
Tasks vs jobs
The best-known study here is “GPTs are GPTs” by Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock (first released in March 2023, later published in Science in 2024). They rated thousands of O*NET tasks and estimated that about 80% of U.S. workers could have at least 10% of their tasks affected by large language models, while roughly 19% could see at least half of their tasks affected.
Read that twice. “Some of your tasks” is the normal case. “Most of your tasks” is a minority.
Real-world usage points the same way. The first Anthropic Economic Index report (February 2025), built on anonymized Claude conversations, found that about 36% of occupations used AI for at least a quarter of their tasks – but only around 4% used it across three-quarters of them. In that first snapshot, 57% of usage looked like augmentation (people working with the model) and 43% like automation. Later editions, published in September 2025, reported the automation share creeping up, which is worth keeping an eye on.
Exposure vs replacement
Here’s the distinction most headlines skip: exposure is not job loss. “Exposed” means AI could technically do part of the work. Whether a job actually disappears depends on cost, regulation, customer trust, how well the tool really performs, and whether employers choose to cut people or ask them to do more.
The International Labour Organization made this point directly in its refined generative AI exposure index (May 2025). It found that roughly one in four workers worldwide is in an occupation with some degree of generative AI exposure, while only about 3.3% sit in the highest-exposure category. Clerical jobs came out as the most exposed group. The ILO’s own conclusion was that transformation of jobs is a far more likely outcome than full automation.
So a high test score is a heads-up, not a forecast.
Why entry-level roles feel it first
This is the part that should get early-career readers’ attention. In August 2025, Erik Brynjolfsson, Bharat Chandar and Ruyu Chen at the Stanford Digital Economy Lab published “Canaries in the Coal Mine?”, using payroll data from ADP. Since late 2022, workers aged 22–25 in the most AI-exposed occupations saw about a 13% relative decline in employment. Older workers in the same occupations held steady or grew.
Why? Junior jobs are packed with exactly the tasks that score high on our test: drafting, summarizing, data cleanup, first-line support. Senior people carry more of the judgment, accountability and client trust – the low-scoring stuff. The study also found the declines were concentrated where AI was used to automate work rather than augment it. We dig into this in are entry-level jobs disappearing.
One more thing: this page is about your current job, not a ranking of occupations. If you want the bigger lists, we’ve covered which jobs AI will replace by 2030 and which jobs AI will replace first separately, and the other end of the spectrum lives in our guide to AI-proof careers.
Warning signs inside your company
Your score tells you what’s technically possible. Your company tells you what’s actually going to happen. If you want a real answer to is my job safe from AI, watch what your employer does, not what it says in the all-hands.
Junior hiring freezes
When people leave and the junior roles behind them quietly aren’t refilled, that’s often the first move. It rarely gets announced. You just notice the team got smaller and nobody’s talking about it.
AI tools rolled out to your team
A new AI assistant on its own is neutral – sometimes it’s simply a perk. The detail to watch is the framing. “Try this to save time” is one thing. “This should let the team absorb the extra volume” is another.
Tasks being consolidated
Three roles merged into one. One person now “overseeing” what four people used to do by hand. Consolidation is usually the stage where automation turns into headcount decisions.
New metrics
Watch for targets that suddenly jump – tickets per hour, reports per week, content pieces per sprint. When the bar rises sharply, it often means someone upstairs has assumed AI will cover the gap.
Here’s a quick checklist to revisit every month or so:
- Open roles on my team are being left empty, especially junior ones
- We got an AI tool, and it came with higher output expectations
- My highest-scoring tasks are the ones the tool is aimed at
- Roles or teams near mine have been merged
- My KPIs changed in the last six months without much explanation
- Leadership talks about “efficiency” more than “growth”
One or two ticks? Normal for 2026. Four or more, combined with a high test score, and it’s time to act rather than wait.
What to do this month, by tier
Thirty days is enough to change how you work – not your whole career, but your trajectory. Pick your column and go.
| Week | Augmented (0–35) | Reshaped (36–55) | At risk (56–80) |
|---|---|---|---|
| Week 1 | Try one AI tool on your worst admin task. Note the time saved. | List your tasks scoring 6+. Test an AI tool on the top two. | Redo the test with a colleague. Confirm which tasks are really exposed. |
| Week 2 | Turn that one win into a repeatable routine or template. | Build a simple workflow: AI drafts, you review and decide. | Learn to do your highest-scoring task with AI, start to finish. |
| Week 3 | Show a teammate what you did. Being the helpful one counts. | Shift an hour a week toward judgment-heavy, low-scoring work. | Research one adjacent role that scores lower. Talk to someone in it. |
| Week 4 | Pick the next task and repeat. | Tell your manager what you’ve changed and what it saved. | Pitch owning the AI version of your process, and start a structured course. |
Whatever column you’re in, the skills that keep their value look similar: knowing how to brief an AI tool clearly, checking its output critically, working with data, and the human stuff – explaining, persuading, taking responsibility. We track what employers are asking for in what AI skills are in demand.
This table is the short version. For the full, step-by-step playbook, read how to not get replaced by AI at work.
Whatever your score, the move is the same
Augmented, reshaped or at risk – the next step points in one direction: become the person on your team who uses AI best. That’s the person who gets asked to design the new workflow instead of being written out of it.
If you’d rather not piece that together from random videos, Coursiv’s 28-Day AI Certificate Program walks you through practical AI tools in short daily lessons, and you finish with a certificate of completion. It won’t guarantee anyone’s job – nothing can – but it gives you a structured way to turn your score into skills.
Want to see what’s covered first? Browse the AI Certificate Program syllabus, then come back and retake the test in a month. We’d bet your number looks different.