Is Your Job Safe From AI? Answer These 5 Questions to Find Out

No job is safe from AI as a whole, and no job is doomed as a whole either. Exposure lives in tasks, not titles. The five questions below sort the automatable part of your week from the part that still needs a person deciding.

Answer them in order and you will know where you stand by the last one.

They run from the work itself, to what the work requires, to who gets hired at the start, to the judgment that earns money, to what your own employer is deploying.

Write the answers down. Vague worry is what makes people freeze, and a written answer is something you can act on this week.

1. How much of your week is routine text?

How much of your week is routine text?

Prompt to copy

List every task I did last week, one line each. Next to each, write just the tag that fits: text, numbers, routine rules, physical action, or live human judgment. Then tell me which three tasks involve the most drafting, summarizing or sorting, and which three needed a person to decide something.

That prompt goes into ChatGPT, Claude or Gemini and takes about 10 to 60 seconds to answer.

The point is the sort: the more marks land on drafting, summarizing, sorting and fixed rules, the more an assistant can speed up your week.

The judgment marks are what stays yours.

The speed is real. A 700-word status page runs about 40 minutes by hand, comes back in 10 to 60 seconds, then takes 15 to 20 minutes checking names, figures and claims.

A 60-page PDF or one-hour transcript takes about two hours to read and note, summarizes in 1 to 2 minutes, then 20 to 30 minutes verifying what the summary got wrong.

Free tiers vary, so check each product's own page before you plan around one. The real cost: about 15 to 20 minutes of your review on a status page, 20 to 30 on a long document, against 40 minutes and two hours by hand.

Name the one task where a draft saves time but you still decide.

💡 Tip: Keep the sort in a plain spreadsheet, one task per row, with a column for minutes spent. It becomes your before-and-after record.

⚠️ COMMON MISTAKE

Never paste private records into a chatbot

On free and personal plans, chats may be used to improve the company's models unless you switch that off in settings. Passwords, client files, and medical or financial records stay out of the box. Every fact, quote and link the assistant produces gets checked before you use it.

2. Does your job need a body, a license, or a live call?

Does your job need a body, a license, or a live call?

A tool can draft the case file, but it cannot stand in the room or carry the license, so that part of the work is not a candidate for handing over.

A job with a body in the room, a license, or a call made in seconds is not immune to AI. It is harder to automate because one person has to answer for the result.

The Resume Now index, reported by Forbes in January 2026, scores the jobs where that duty sits. Nurse anesthetists score 93.3, at a median $195,263.

Emergency physicians score 92.3, at $302,047. Judges and general surgeons both score 91.3, at $115,325 and $339,027. Commercial pilots score 91.0, at $101,876.

Read what that index counts before you trust it. It only includes jobs scoring 74 or higher on adaptability, stress tolerance and self-control, with median pay of at least $74,000. Those are traits and wages. It is not a test of what software can do.

Anthropic's March 2026 report looks at the same question from the other end. It puts 30% of jobs at almost no risk of AI elimination. Cooks, motorcycle mechanics, lifeguards, bartenders and dishwashers are on that list.

The catch is that a license is not a shield over the paperwork. AI still writes clinical notes, case summaries and customer messages in hospitals, cockpits and courts. Your sign-off stays yours.

The filing around it does not. Name the part of your job that needs your presence or your name on the line, and build around it.

3. Does your career stage change the risk?

Does your career stage change the risk?

It does, in one direction. The first rung of a career is where AI does the most damage.

Jobs with a body in the room or a name on a license keep their older workers well.

They say nothing about who gets hired at 23, and that is where the squeeze is.

Stanford HAI's Canaries Dashboard, October 2026, tracked 22- to 25-year-olds in software development and customer service.

Their employment has dropped over 3% a year since ChatGPT launched in November 2022, and 4.4% in recent months. The bottom rung is narrowing.

The whole occupation is not vanishing.

The Cleveland Fed, in November 2026, put numbers on the age gap. For workers 23 to 30, the highest AI exposure raises predicted unemployment by about two points. For workers 45 and older, it is at most one point.

The steady headcount at your employer proves little. In exposed jobs, people still leave at the same rate, but fewer get hired. A team that looks stable can be a closed door.

So aim at work that shows senior judgment. Own a client outcome, lead a small project, write down the decisions you made and why. Tool familiarity alone is not the edge.

By the numbers: 40 minutes to read and note a 60-page file by hand against 1 to 2 minutes in Claude, plus 20 to 30 minutes of checking. The tool saves reading time.

💡 Tip: Count the roles your team has posted in the last six months and whether each was filled. That count is more informative than any national headline.

What Actually Protects You

✅

Builds your footing

  • Owning a client outcome end to end
  • Leading a small project and being named on it
  • Writing down why you overrode a draft or a model's answer
  • Explaining the problem and the context before asking for help
  • A dated record of what changed at work
🚫

Does not, on its own

  • Knowing your way around five different AI tools
  • Having a license with no sign-off work behind it
  • Keeping your head down through a quiet hiring market
  • Assuming a stable team means a stable next move
  • Waiting for a national report to tell you about your own workplace

4. What can you show that AI cannot do without you?

What can you show that AI cannot do without you?

Correcting a bot's draft is the dull part, and it may be the best proof you have.

A study from Wharton Human-AI Research in November 2026 said staff need to state the problem, give the background, question the answers and judge if the output fits.

Knowing the tools alone is not enough.

PwC's 2026 Global AI Jobs Barometer read more than a billion job ads and found AI-exposed roles seven times more likely to ask for leadership, creativity and decision-making. Entry-level ads rebuilt around senior skills are up 35% since 2019; the old kind is down 10%.

So write down one recent case where you fixed an AI draft, in three lines: the situation, what the tool produced, and what you changed and why. That note answers the question, and it is the same proof those senior postings want.

💡 Tip: Keep three of these examples in a single document, each dated. Refresh it after every project rather than the night before a review.

🗓️ Your 90-Day Test, In Order

1

Week 1 – Write the answers

Put your five answers in one dated document so the starting point is fixed.

2

Week 2 – Ask for one task

Request a single piece of work with more judgment or client ownership attached to it.

3

Weeks 3 to 8 – Keep the record

Note what you decided, what you overrode and what changed on your team, with dates.

4

Week 9 – Re-read the five answers

Compare what you wrote in week 1 against what your job actually looks like now.

5

Week 12 – Decide the next move

Pick upskilling, a role switch or staying put, and write down which answer drove it.

5. What is your employer actually deploying?

What is your employer actually deploying?

Strong judgment is worth little if your team never sees it used, so check the one workplace before you act on national headlines. Those numbers describe a labor market, not your team, and fear alone is the wrong reason to move.

Check three signals: internal AI tools you are asked to use, postings for your role, and whether recent departures were backfilled.

The AIimpacted ledger, snapshot 5 October 2026, lists 42 employer-attributed AI job-loss events totaling 81,486 jobs. Real cuts exist, but that is not proof about your team.

Prompt to copy

Act as my workplace analyst. Here are three signals I observed: [internal AI tools rolled out or requested], [job postings and role wording I saw], [departures and whether they were backfilled]. Summarize each signal in two sentences, tell me whether they point to task change or role change, list what evidence is missing, and draft four questions I can ask my manager. Keep it under 350 words and do not speculate beyond the observations.

A chat assistant returns that summary in under a minute, and the review takes about 15 minutes. Then set a 90-day review date, ask for one task with more judgment or client ownership, and keep a dated record to compare against.

Frequently Asked Questions

Do I need to code to run this self-check?

No. Every step uses a chat assistant you already type into, a spreadsheet, or a conversation with your manager. The only technical skill is copying and pasting your own task list and correcting what comes back.

Will answering these five questions tell me if I will lose my job?

No, and anyone who promises that is guessing. The questions tell you which tasks an AI can already draft and which still need a person to decide. That is the part you can act on, and it is more useful than a prediction.

What does it cost to do this?

The free tiers of ChatGPT, Claude, Gemini, Microsoft Copilot and Perplexity are enough for the drafting and summarizing in question one, and their usage limits change, so check each product's own page. Paid plans only matter if you hit those limits often.

Can I paste my work documents into the assistant?

Not client files, passwords, or medical and financial records. On free and personal plans, chats may be used to improve the company's models unless you switch that off in settings. Paste a description of the document instead, or strip names and figures first.

How long should this take from start to finish?

Question one takes about an hour, mostly reading your own calendar and email. Questions two through four take 20 minutes each. Question five takes an afternoon of asking and one 90-day wait before you compare the answers.

My answers came out mixed. What now?

Fix the weakest of the five first, which is usually question five. Most people can control what their employer is deploying far more than they can control a national trend, and the 90-day record costs almost nothing to start.

Your First Task To Defend

If your answers cluster on text-heavy, routine, early-career work, treat the role as exposed. Chase one task with senior judgment attached this month, and write down why you made the calls you made.

If they cluster on licensed, physical or real-time work, use AI to cut the documentation around the job and protect the sign-off. If the answers are mixed, fix the weakest of the five before touching the rest.

This week, act on the answer to question five. Ask your manager which of your tasks are being reassigned and who picks them up, then put the date in your calendar.

Leave a Comment