Your Job Is Not Disappearing. Your Tasks Are.

Every job is a bundle. A junior accountant does data entry, reconciliation, drafting, checking, explaining to a client, and staying late in March. A designer does research, sketching, production, revisions, file handover and managing a nervous stakeholder.

Automation has never taken a bundle. It takes items out of the bundle. The bundle then gets refilled with something else — usually more of whatever the machine cannot do.

This is not optimism. It is just how the last two hundred years of this have gone, and it is a more useful frame than the headline version, because it tells you what to do on Monday.

How to tell which of your tasks are at risk

Go through your week and score each task on three questions. The more yeses, the sooner it goes.

• Is it repetitive? Does it look roughly the same every time?
• Is it written down? Could a competent stranger follow instructions and do it?
• Is a wrong answer catchable? Would you notice a mistake quickly and cheaply?

Three yeses means it is already being automated somewhere. Formatting reports. First-draft copy. Reconciling two lists. Categorising tickets. Extracting numbers from documents.

Now look at the tasks with three noes. Deciding what to do when the rule and the sensible answer disagree. Telling a client bad news. Knowing that the number is technically right but obviously wrong. Choosing what not to build. These are not safe forever, but they are safe for a long while, because they depend on context nobody has written down and on someone being accountable.

The question is not “can a machine do my job”. It is “what percentage of my week is machine-shaped, and what am I doing with the rest”.

The part nobody wants to say out loud

Here is the genuine problem, and it is not mass unemployment.

The tasks disappearing first are, almost exactly, the tasks juniors used to do. That is not a coincidence — work gets given to juniors precisely because it is repetitive, documented and checkable.

Which means the ladder loses its bottom rung. Experienced people get more productive. New people get fewer ways in. A senior analyst with an agent is very valuable. The way you became a senior analyst was by spending two years doing the thing the agent now does.

Nobody has solved this yet. The companies that will solve it are the ones that deliberately teach the judgement layer instead of waiting for it to be absorbed by osmosis from grunt work.

What to do if you are early

Get to the review layer fast. Reviewing output — spotting what is wrong, knowing what “good” looks like, handling the exception the system could not — is the job that is growing. You can learn it faster than the old apprenticeship allowed, because you can review a hundred outputs in the time it used to take to produce three.

Also: get close to the customer. Anything that involves understanding what a person actually wants, as opposed to what they asked for, stays human for a long time.

What to do if you are experienced

Your risk is different. It is not that you get replaced; it is that you stay expensive while doing the tasks that got cheap. Audit your own week honestly. If a third of it is machine-shaped, that third is where your next review conversation goes.

The upside is real, though. Experience is exactly what makes an agent useful, because you can tell when it is wrong. That skill has no shortcut and it just became more valuable, not less.

A note on titles

Job titles are lagging indicators. They change years after the work does. Do not wait for your title to tell you the bundle has shifted — by then everyone else has already noticed.