Will AI Replace My Job? Task Exposure Calculator

1. Split a normal work week into the six task groups and enter the hours for each. Leave a group blank if you never do it.
2. Give each group a rating: E0, E1 or E2. Rate what an AI tool could do today, not what you fear it could do one day.
3. Read the middle estimate, then the lower and upper bounds. Change one rating to see how much it moves the result.

The three ratings come from a published study. Use these plain meanings:

What each result tells you:

AI job exposure is the share of the tasks in a job that an AI tool could complete much faster at the same quality. It describes tasks, not people. It is not the same as job loss.

The measure used here comes from the study "GPTs are GPTs" by Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock. The study rated work tasks from the O*NET database of United States occupations. A task counted as exposed when a large language model could cut the time to complete it by at least 50 percent at equal quality.

The authors found that around 80 percent of the United States workforce could have at least 10 percent of their work tasks affected. About 19 percent of workers may see at least 50 percent of their tasks affected. The authors state that they make no predictions about the timeline of development or adoption.

That last point matters. Exposure answers "could a tool speed this task up?". It does not answer "will my employer cut this role?". This page cannot answer the second question, and no honest tool can.

The calculator adds up the hours you rated E1 and E2 and divides by your total hours. It reports three measures, the same three the study defines.

1. Total hours = the sum of the hours in all six task groups.
2. E1 hours = hours in groups you rated E1. E2 hours = hours in groups you rated E2.
3. The lower bound counts E1 only. The study calls this the lower bound of exposed tasks.
4. The middle estimate gives E2 half weight. The study uses a 0.5 weight because extra tools need more investment before they work.
5. The upper bound counts E1 and E2 in full. The study calls this the upper bound of exposure.

One thing differs from the study. The study counts tasks within an occupation. This tool weights by your hours, because hours describe your own week better than a task list does. Your result is therefore not directly comparable with the study percentages.

This is an illustration with made-up ratings. It is not a finding about any real job. A desk worker spends a 40-hour week like this:

Meaning: by these ratings, 23 of 40 weekly hours sit in tasks a tool could speed up. The other hours do not. A hands-on week with 25 of 40 hours rated E0 gives a middle estimate of 15.0 percent with the ratings in Example B.

Your result depends on your inputs alone. Four things move it most.

How You Split Your Hours

The hours are the weights. Moving 5 hours from an E0 group to an E1 group in a 40-hour week raises the lower bound by 12.5 points.

How Strictly You Rate

The study sets a high bar: the time must fall by at least half at equal quality. A tool that helps a little is still E0. Loose ratings inflate the result.

E1 or E2

E2 hours count half in the middle estimate. Rating 10 hours as E2 instead of E1 lowers the middle estimate by 12.5 points in a 40-hour week.

How Mixed Each Group Is

A group holds many tasks. Rate the group by what fills most of its hours. For a closer result, look up the task list for your occupation in O*NET OnLine and rate each task.

When You Rate

Tools change. A rating made today describes today. Run the tool again when your tools or your duties change.

Exposure and displacement are different things. Most scare headlines mix them up.

The International Labour Organization made the same point in its 2025 global index. It found that one in four workers worldwide are in an occupation with some generative AI exposure, and that 3.3 percent of global employment falls in the highest exposure category. It concluded that transformation of jobs is the most likely impact, because most occupations include tasks that need human input.

The same report found higher exposure in high-income countries, at 34 percent of employment, than in low-income countries, at 11 percent. Clerical occupations had the highest exposure.

When You Plan Your Skills

Find the task group that adds the most exposed hours. Then decide whether to learn the tools for it or to shift time toward tasks you rated E0. Before paying for training, test the payback with the bootcamp ROI calculator or the college ROI calculator.

When You Compare Two Roles

Run the tool once for each role with honest hours. The guide on a raise versus a promotion helps with the pay side of that choice.

When You Want to Save Time Now

Exposed hours are also hours you could speed up yourself. The AI workflow efficiency calculator puts a time value on that.

When a Headline Worries You

Replace a vague fear with your own numbers. A result of 30 percent means 70 percent of your week is unchanged by your own ratings.

1. Treating Exposure as a Layoff Forecast

The study behind this rubric makes no prediction about timing or adoption. An exposed task can make a worker faster instead of redundant.

2. Trusting a Single Risk Score for a Job Title

Two people with the same title can have different weeks. A score that ignores your own task mix cannot describe you.

3. Rating by Fear

Rate what a tool does today at equal quality. Rating every task E1 because AI "might" do it one day turns the result into a guess.

4. Ignoring the Quality Test

The rubric needs equal quality. A fast draft that you must rewrite does not halve the time.

5. Comparing Your Hours Result with Study Percentages

The study counts tasks across occupations. This tool weights your own hours. The two numbers measure different things.

6. Believing Exact Timelines

We found no cited study that gives a reliable date for one person or one job. Be careful with any tool that prints one.

The arithmetic is exact. The result is only as good as your ratings, and it says nothing about whether your job will change.

The rubric dates from 2023 and was written for large language models. Your ratings should reflect the tools you can use today.

Nobody can answer that for one person with a calculator. Research measures exposure, meaning tasks a tool could speed up. The International Labour Organization concluded in 2025 that transformation of jobs is the most likely impact of generative AI.

It means 40 percent of your weekly hours sit in tasks you rated as ones an AI tool could halve. In a 40-hour week that is 16 hours. It does not mean a 40 percent chance of losing your job.

Clerical occupations have the highest exposure in the 2025 International Labour Organization index. The 2023 study behind this tool found that higher-income jobs potentially face greater exposure to language models.

One 2023 study estimated that around 80 percent of the United States workforce could have at least 10 percent of their tasks affected, and about 19 percent could have at least 50 percent affected. A 2025 global index put one in four workers in an occupation with some exposure.

E1 means a chat-style AI language tool alone could halve the time for the task. E2 means it would take extra software built on that tool to halve the time. E2 counts half in the middle estimate.

A job title would need a built-in risk score, and we have no sourced score for individual titles. Your own task hours describe your week better. You can look up the official task list for your occupation in O*NET OnLine.

This page gives no timeline. The authors of the study behind the rubric state that they make no predictions about the development or adoption timeline. Adoption depends on cost, rules and employers.

No. An exposed task is one you could also finish faster yourself. The study concludes that these tools show traits of general-purpose technologies, which change how work is done across many fields.

Anyone making a major decision, such as quitting a job or paying for a degree, should not rely on it alone. It reflects your own ratings on one day and says nothing about hiring or pay.

No. The calculation runs in your browser and nothing is sent to our servers. Anything you Save stays in this browser only.

• GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models (Eloundou, Manning, Mishkin and Rock, arXiv, version 5, 21 August 2023: the E0, E1, E2 rubric, the three measures and the 80 and 19 percent findings).
• Generative AI and Jobs: A Refined Global Index of Occupational Exposure (International Labour Organization, Working Paper 140, 20 May 2025: global exposure shares).
• O*NET OnLine (U.S. Department of Labor sponsored database of occupations and their tasks, the task source used by the 2023 study).