The short answer
Nobody can tell you whether your specific job survives. But you can judge your own exposure by looking at which tasks you do, not which title you hold.
- Producing a competent first draft of routine writing.
- Summarising, reformatting, and translating between styles or languages.
- Extracting structure from messy documents.
- Generating many options quickly when you want quantity over precision.
- Answering well-covered questions in domains with abundant written material.
Nobody can honestly tell you whether your particular job survives, and anyone who claims certainty in either direction is selling something. What you can do is assess your own exposure sensibly, and the key move is to stop thinking about job titles and start thinking about tasks.
Almost no job is a single activity. It is a bundle of a dozen different things, and those things face very different levels of pressure. Some of what you do may be highly exposed while the rest is barely touched. That mix, not your title, is what determines your position.
Break your job into tasks first
Write down what you actually spent your last two working weeks doing. Not your job description. The real list, including the parts nobody put in the description.
Then sort each item by two questions. Is this mainly producing or processing text, code, images, or structured data? And is a roughly-right answer acceptable, or does it have to be correct, defensible, and owned by someone?
Tasks that are text-shaped and tolerant of a first draft are the exposed ones. Tasks requiring physical presence, accountability, negotiation, judgement under ambiguity, or relationships built over time are far less so.
What these systems are genuinely good at
Being specific here is more useful than generalities about capability.
- Producing a competent first draft of routine writing.
- Summarising, reformatting, and translating between styles or languages.
- Extracting structure from messy documents.
- Generating many options quickly when you want quantity over precision.
- Answering well-covered questions in domains with abundant written material.
- Writing and explaining code, especially familiar patterns.
The pattern is transformation of information that already exists, where a human checks the result. That is a real capability and it covers a substantial slice of white-collar work.
What they remain weak at
The limitations are equally specific and mostly durable.
- Being accountable. Someone has to sign off, absorb the consequences, and answer for the decision.
- Operating with missing information, where the skill is knowing which question to ask.
- Reading a room. Sensing that the stated problem is not the real one.
- Anything physical, unstructured, and variable. Robotics has advanced far more slowly than text generation.
- Sustained relationships and trust built over years.
- Knowing what is worth doing at all, rather than doing what was asked well.
Replacement is not the usual mechanism
The popular image is a job being eliminated and a person shown the door. That happens, but the more common effects are subtler and easier to miss while they are happening to you.
Roles get redefined. The drafting portion shrinks, the reviewing and directing portion grows. The work changes character even though the title stays.
Hiring slows quietly. Rather than making cuts, an organisation replaces fewer people who leave. From inside, nothing dramatic occurs. The team just gets thinner and stays that way.
Entry-level rungs erode. This is the effect worth watching most closely. A lot of junior work has historically been the routine production that senior people did not want to do, and that is precisely the exposed category. If the training ground disappears, the pipeline into experienced roles gets harder to climb, which is a problem for entire professions rather than individuals.
Expectations rise. When drafts are cheap, the standard for what counts as finished work moves up. The volume expected of you increases with the tools available.
Why economists disagree
This is a live argument among serious people, and it is worth understanding why rather than picking whichever forecast you find comforting.
One camp points out that automation has repeatedly displaced specific occupations while creating new demand elsewhere, often in roles nobody could have predicted. On this view, the fear is real for individuals but the aggregate employment picture works out.
Another camp argues this wave differs in an important way. Previous automation mostly hit physical and routine tasks, pushing workers toward cognitive ones. This one hits cognitive tasks directly, which narrows the obvious place to move to.
A third position focuses less on whether jobs exist and more on how the gains get distributed. Even if employment holds up, the value may concentrate in ways that leave a lot of people worse off.
All three have credible arguments. The honest summary is that the direction of change is clearer than its magnitude, and the timeline is genuinely uncertain. Forecasts in this area have a poor track record in both directions.
What actually improves your position
Given real uncertainty, the sensible strategy is the one that pays off across several possible futures rather than betting on one.
Get fluent with the tools. The near-term displacement pressure comes far more from colleagues using these systems well than from the systems themselves. Fluency is cheap to acquire and immediately useful.
Move toward judgement and ownership. Within your existing role, take on more of the work that involves deciding, defending, and being answerable. That is the part that does not delegate cleanly.
Build the things that do not transfer. Domain knowledge specific to your organisation, relationships with people who trust you, an understanding of how decisions actually get made where you work. None of this is in any training corpus.
Combine skills rather than deepening only one. The intersections are harder to automate than any single competency. Someone who understands both the technical side and the customer side is filling a gap that a tool cannot see.
Stay close to physical or interpersonal reality. Work involving hands, bodies, unpredictable environments, or care remains substantially harder to automate than the discourse suggests.
A realistic assessment
If most of your working day is producing routine text or code that someone reviews lightly, you should take this seriously and start shifting the mix of what you do. Not in a panic, but deliberately, over the next year or two.
If your work involves physical presence, real accountability, ambiguous problems, or trust built over time, the immediate risk is lower. The likely change is that parts of your job get faster, and that expectations adjust accordingly.
Either way, the useful posture is neither dismissal nor dread. Something real is happening to the composition of a lot of jobs, the pace is uncertain, and the people who fare best will mostly be the ones who noticed early which parts of their own work were the exposed parts.
Comments
0 total