Zrooth
Menu

Blog

Which Parts of Your Job Are Most at Risk from AI? A Practical Self-Check

18 November 2026 · 8 min read

Worker reviewing a task worksheet at a desk, symbolizing self-checking personal ai job risk

The parts of your job most at risk from AI are particular tasks, not your job title: routine work in text, numbers or code, with a clear right answer, that repeats every week and that nobody has to sign for. The quickest way to measure your own AI job risk is to list your tasks, score each one on five questions, and see how many of your hours fall into the high-scoring group.

This article is a worksheet, not a forecast. A companion piece, Which Skills AI Makes Obsolete, and Which It Rewards, describes the two groups of skills in general. Here you apply the same thinking to your own working week, with a pen, a spreadsheet and about an hour.

Why Score Tasks Instead of Your Job Title?

Because AI does not arrive one job at a time. It arrives one task at a time, and every job is a bundle of tasks with very different exposure. The International Labour Organization's 2025 refined index of generative AI exposure, which scores occupations at the task level, found that about one in four workers worldwide is in an occupation with some exposure, but only 3.3% of global employment falls into the highest exposure category. Its conclusion is that, because most occupations contain tasks that need human input, transformation of jobs is the most likely impact.

Usage data points the same way. Anthropic's Economic Index, published in February 2025, mapped millions of conversations with its Claude assistant to work tasks and found that roughly 36% of occupations used AI for at least a quarter of their tasks, while only about 4% used it for three quarters or more. It also found that 57% of the usage looked like augmentation, people working with the AI, against 43% automation. In plain terms: your title tells you little. The mix of tasks inside it tells you a lot.

Step One: List Your Tasks for One Real Week

Take a recent, ordinary week and write down every distinct task you did, with the hours each took. Aim for ten to fifteen lines. Be specific: "reply to vendor queries about payment status" is a task; "accounts" is not.

  • Use your calendar and sent mail as evidence, not your job description. Job descriptions describe the role as it was imagined.
  • Include the invisible work: chasing people, fixing other people's errors, preparing for meetings, training a new colleague.
  • If you are stuck, look up your occupation on O*NET OnLine, run by the US Department of Labor, which holds more than 19,000 occupation-specific task statements. Use it to jog your memory, then go back to your own week, which is the only list that counts.

Step Two: Score Each Task on Five Questions

Score every task from 0 to 2 on each of the five questions below. Add the five scores to give each task a total between 0 and 10.

Question0 points1 point2 points
1. Are the inputs and outputs text, numbers or code?Mostly physical or face to faceMixedAlmost entirely digital
2. Is there a clear right answer someone can check quickly?Judgement, no single answerPartly checkableClearly right or wrong
3. Does it repeat in a similar form?Rare or one-offMonthlyWeekly or daily
4. Can it be done without knowing the specific people involved?Depends on relationships and historySome context neededAnyone with the inputs could do it
5. Does nobody have to sign for or answer for the result?You carry legal or financial responsibilitySomeone reviews itNobody is accountable for it

How to Read Each Task's Score

  • 0 to 3, low exposure: work that depends on judgement, relationships or responsibility. AI may help you prepare it, but the task itself stays with a person.
  • 4 to 6, changing: AI will reshape how the task is done. Expect to do it faster, with AI drafting and you checking.
  • 7 to 10, high exposure: tools can already do much of this, or soon will. Plan on these hours shrinking.

These bands are a practical rule of thumb, not a scientific measure. The questions follow the pattern the research above describes: digital, checkable, repetitive work without accountability is where AI use concentrates.

Step Three: Work Out Your AI Job Risk Share

Now weight by time. Add up the hours spent on tasks scoring 7 or more, and divide by your total hours. That percentage is your AI job risk share: the part of your week most likely to be done differently, or by a tool, within a few years.

An Illustrative Example

Here is a hypothetical accounts executive at a trading company in Navi Mumbai, working a 40-hour week. The scores are illustrative, not a measured case.

TaskHoursScore
Entering supplier invoices into the accounting system109
Bank reconciliation68
Preparing the month-end summary report57
Answering vendor emails about payment status66
Resolving disputed invoices with long-standing suppliers53
Preparing documents for the auditor and answering queries42
Training a new junior colleague41

The three high-scoring tasks take 21 of 40 hours, a share of about 52%. That looks alarming, but read it properly: it says half of this person's current week is the part to redesign, and the other half (disputes, audit, training) is where their value will increasingly sit.

What Does a High Score Actually Mean?

A high score means exposure, not dismissal. Whether exposed hours turn into fewer jobs, different jobs or more output from the same people depends on the employer, the market and what you do next.

One finding is worth taking seriously, though. In a study of 5,179 customer support agents, Brynjolfsson, Li and Raymond found that a generative AI assistant raised productivity by about 14% on average, and by 34% for novice and lower-skilled workers, with minimal impact on the most experienced. If your edge in a task is simply that you have done it for years, AI may narrow that edge, because it lifts newcomers closest to the experienced level.

The flip side is checking. In Stack Overflow's 2025 developer survey, 66% of developers named AI solutions that are "almost right, but not quite" as their biggest frustration. Someone has to catch the "not quite". In a high-scoring task, the person who can check, correct and own the output is worth more than the person who only produces it.

What to Do With Your Results

  1. For tasks scoring 7 to 10: learn to do them with AI now, and offer to redesign the process yourself. The person who maps the work and builds the fix is rarely the person whose hours disappear. Mapping where a business leaks time and money is a good place to start.
  2. For tasks scoring 4 to 6: practise the checking. Pick one task, try a tool on a real example, and judge the result honestly, as described in judging an AI tool against a real task.
  3. For tasks scoring 0 to 3: grow them on purpose. Ask for more of the work that needs relationships, judgement and responsibility, and make it visible to your manager.
  4. Decide what to delegate: some exposed tasks can go to an AI agent with guardrails, others should not. See what you can and cannot delegate to an agent.
  5. Repeat the check every six months. Tools change quickly, and a task scoring 5 today may score 8 next year.

A Note for India, the UAE and Kenya

Exposure is not spread evenly. The ILO index puts overall exposure at 34% of employment in the richest group of countries and 11% in the poorest group, and finds that 4.7% of women's employment sits in the highest exposure category against 2.4% for men, largely because clerical work is so exposed. For the large back-office, shared-services and customer operations workforces in Indian cities, Dubai and Nairobi, that means the self-check matters more, not less: much of this work is digital, repetitive and checkable by design.

Common Mistakes When Scoring Yourself

  • Scoring the title, not the week. "Analyst" tells you nothing. The individual tasks tell you everything.
  • Being too kind. Most people underrate how checkable their own work is. Ask a colleague to score your list too.
  • Ignoring hours. One high-scoring task that takes half your week matters more than five that take an hour each.
  • Treating the score as fate. The number describes your current week. You can change the mix, and that is the point of measuring it.

Turning the Self-Check Into a Plan

A self-check is only useful if it changes what you do on Monday. If your AI job risk share is high and you want structured practice redesigning that kind of work, the AI Entrepreneurship Programme spends twelve weeks applying AI to how real businesses run: mapping processes, judging tools and building working fixes. What you learn sets out the sequence, and Pathfinder shows how it applies to your industry. If you have already lost a role, how to rethink your skills after an AI layoff may be the better next read.

The programme is business consulting and training. It does not promise a job, income or any particular outcome; it teaches a method you can apply to your own work.

Sources

Frequently Asked Questions

How do I know which parts of my job are at risk from AI?

List the tasks from one real week with the hours each takes, score each on five questions about how digital, checkable, repetitive, context-free and unaccountable it is, then add up the hours on high-scoring tasks.

Is a high AI job risk score a sign I will lose my job?

No. It measures exposure, not dismissal. Whether exposed hours become fewer jobs, different jobs or more output depends on your employer, your market and how you change the mix of your work.

Which tasks are least exposed to AI?

Tasks that depend on relationships, judgement about specific people or situations, physical presence, or legal and financial responsibility. AI can help prepare them, but a person still does and answers for them.

How often should I repeat the self-check?

Every six months is a sensible rhythm. Tools improve quickly, so a task that scores in the middle today may move into the high band within a year.

Does the self-check work for technical roles such as software developers?

Yes. Developers should list tasks such as writing boilerplate, reviewing code, debugging, talking to users and making design decisions separately, because their exposure differs widely even within one role.

Where can I find a list of typical tasks for my occupation?

O*NET OnLine, run by the US Department of Labor, holds over 19,000 occupation-specific task statements. Use it to prompt your memory, then rely on what you actually did in a real week.

What should I do first if most of my hours score high?

Pick the biggest high-scoring task, learn to do it with an AI tool on a real example, check the output carefully, and offer to redesign that process yourself rather than waiting for someone else to.

Zrooth Mirror · Free · 12 Minutes

Where Do You Really Stand in Your Career?

Not a flattering quiz. An honest look at the gap between what you want and what you actually do, with one clear next step. Then watch the Zrooth Mirror Series.

Pathfinder · Free

Where Do You Fit in the AI Economy?

Type your industry and the work you do. See the live businesses you would learn inside, your twelve weeks, and three venture ideas made for you.

Find My Path

Join the Discussion

Agree, disagree or add your own experience. Every comment is reviewed before it appears; spam and abuse are removed, and disagreement is welcome.

    Add Your View

    0 / 2,000
    Links send a comment to manual review. Please do not share personal details about other people.