Number AI Job Risk Calculator
AI does not replace jobs; it reaches tasks. So this calculator never asks for your job title alone — it scores the mix of tasks your week is actually made of, shows its weights openly, and tells you which of your tasks carry the exposure and which anchor you. Pick a preset to start, then adjust the mix until it looks like your real week.
Your task breakdown
| Task | Share of week | Exposure weight | Contribution |
|---|---|---|---|
| Calculate to see which of your tasks carry the exposure. | |||
Contribution = share × weight. The biggest contribution line is where your exposure actually lives — that is the task to redesign, supervise or move up from, not necessarily the job to leave.
How it works
This is a task-exposure model, in the spirit of the published research that scores occupations by the tasks inside them (the line of work running from Frey & Osborne through the ILO and OECD exposure studies to today’s O*NET-based analyses). Eight kinds of task each carry a fixed exposure weight from 0 to 10 — how directly current AI systems perform or draft that kind of work: routine data handling 9, writing & drafting 8, coding & technical production 8, analysis & reporting 7, customer conversation 6, creative & design work 5, management & judgment 3, physical & manual work 1. Your score is the weighted average of your week, expressed as a percentage and banded: under 25% Low, 25–50% Moderate, 50–75% High, 75%+ Very high. The presets are editorial estimates of each occupation’s typical task mix, set from the qualitative pattern those studies agree on (screen-based routine work most exposed; physical, in-person and care work least) — they are not measurements of any real person’s job. And the honest limit of the whole method: exposure is not destiny. A high score means AI can draft or perform much of your current task mix — in most occupations that has historically meant the job changes shape (more checking, exception-handling and judgment per hour) rather than vanishing, and it says nothing about your employer, your industry’s demand, or you.
Method & worked example
Method: Your score is the weighted average of your week: add up (share of week × exposure weight) across the eight task kinds and divide by the total share — the answer is a 0–100% exposure estimate, banded Low (under 25%), Moderate (25–50%), High (50–75%) and Very high (75%+). The weights are fixed and published on the page (routine data handling 9 down to physical work 1), so any two people with the same mix always get the same score.
Worked example: The data entry preset — 40% routine data handling, 20% writing, 20% analysis, 20% customer conversation — scores 78%, Very high, because 80% of that week sits in the three heaviest-weighted task kinds. The plumber preset — 80% physical work, 10% analysis, 10% customer conversation — scores 21%, Low. Same formula, opposite weeks: the mix is the message.
When to use this
- A headline says your occupation is 'at risk' and you want a number based on your actual week instead of your job title.
- You are choosing where to aim professional development — the breakdown shows which of your tasks to move up from first.
- You manage a team and want a structured way to talk about which tasks AI should take over and which stay human.
- You are picking between two job offers and want to compare the task mixes, not the titles.
How to use it, step by step
- Pick the closest occupation preset, or leave it on Custom — presets only fill the sliders; the sliders are the real input.
- Adjust each slider to the share of your actual week in that kind of task. Be honest about the routine work; everyone underestimates it.
- Read the score and band at the top, then the breakdown table: contributions are ranked, so your biggest exposure is first.
- Try moving 10–20 points of week from your heaviest task toward judgment, creative or physical work and watch the score — that gap is your redesign target.
- Re-run it for a colleague's or a job offer's task mix to compare like with like.
What the inputs mean
| Input | What it is | Typical values |
|---|---|---|
| Occupation preset | A starting task mix for 14 common jobs | Editorial estimates — adjust them to your real week |
| Task shares (8 sliders) | Percent of your week in each kind of task | Normalised automatically; need not total 100 |
| Exposure weights | How directly current AI performs each task kind, 0–10 | Fixed and published on the page — routine data 9, physical 1 |
Another worked example
Two people, same employer, opposite weeks. The office administrator preset (35% routine data, 25% writing, 10% analysis, 20% customer conversation, 10% judgment) scores 74% — High — because 90% of that week sits in tasks weighted 6 or more. The nurse preset (15% routine data, 10% writing, 5% analysis, 25% customer conversation, 10% judgment, 35% physical care work) scores 47% — Moderate — and almost half of that score comes from the documentation quarter of the week, not the hands-on 70% of it. Same building, same 'healthcare/admin' industry label; completely different exposure, because the weeks are different.
Frequently asked questions
My score is High. Should I quit my job?
No — and the model does not say that. High exposure means much of your current task list can be drafted or done by AI tools today. The workers who come out ahead in the research are usually the ones who take over the supervision layer: checking outputs, handling the exceptions, owning the client relationship. The breakdown table shows exactly which tasks to move up from — start there, not with a resignation letter.
Why do two people with the same job title get different scores?
Because the title is not the job — the task mix is. Two bookkeepers can score 40 points apart if one mostly keys invoices (routine data handling, weight 9) and the other mostly advises clients and resolves messes (judgment and conversation, weights 3 and 6). That is precisely why this calculator scores tasks instead of titles, and why the presets are only starting points.
Where do the presets come from?
They are editorial task-mix estimates for typical versions of each job, set to match the qualitative ordering the published exposure research agrees on — data entry, telemarketing and routine bookkeeping at the exposed end; trades, nursing and childcare at the sheltered end. They are deliberately simple, and the page shows every weight so you can adjust them. They are not O*NET data and not a survey of real workplaces.
What actually lowers a score in real life?
Shifting your week’s centre of gravity: less first-draft production (writing, routine analysis, data movement) and more of what AI cannot sign for — physical presence, licensed responsibility, negotiation, care, taste, and being the human a customer or patient trusts. In slider terms, every 10 points of week moved from routine data to judgment work cuts this model’s score by about 6 points.
Is a Low score a guarantee of safety?
No. A low score says today’s AI reaches little of your current task mix. It says nothing about your industry’s economics — a sheltered task inside a shrinking industry is not a safe job, and an exposed task inside a growing one often just becomes a faster job. Treat the number as one lens, not a verdict.
💼 Exposed tasks, valuable human?
If your edge is judgment and clients, price it properly: the Freelance Rate Calculator works backwards from the income you want to the rate you must charge.
Why is customer conversation weighted 6 and not higher? Chatbots answer customers all day now.
They answer the scripted middle of customer work: tracking questions, password resets, standard complaints. The weight sits at 6 — substantial, not total — because a large share of real customer conversation is de-escalation, negotiation, exceptions and trust, which is where current systems are weakest and where liability sits. If your customer work is almost entirely scripted Q&A, raise that slider's share; the model will do the rest.
Can I use this for my whole team or department?
Yes — run it once per role with that role's typical mix, and compare the breakdowns. The useful output for a manager is not the ranking of people but the shared rows: if every role shows routine data handling as its top contribution, that is one automation project that lowers exposure across the whole team at once.