The AI jobs debate runs on a loop. One report declares that AI is creating more roles than it eliminates. The next warns of mass unemployment. Both get shared as proof of whatever the sharer already believed, and both are answering the wrong question. AI is not mainly adding or subtracting jobs. It is changing the shape of the work inside the jobs that remain.
The work changes faster than the headcount
Look at almost any role AI has touched and you see the same thing: the tedious, repeatable core gets automated, and the human is left with the parts that need judgment, taste, or a relationship. The developer spends less time typing boilerplate and more time deciding what should exist and reviewing what the machine produced. The support agent stops answering the same question forty times and handles the genuinely hard cases. The headcount may barely move while the actual day is transformed. Averaged into a single number, that shows up as either “jobs created” or “jobs lost” depending on who is counting, which is why the reports contradict each other.
The real divide is inside every role
The meaningful split is not between industries, it is between the loop-shaped part of a job and the judgment-shaped part. If a role is almost entirely a loop, running the same steps on new inputs, it is exposed, and pretending otherwise helps no one. If a role is mostly judgment, it gets amplified, because the person can now offload the grind and spend their attention where it is scarce. Most real jobs are a mix, and AI is quietly moving the mix toward judgment. That is the same reason we tell owners to map the loop before they hire: the loop is what the machine takes.
What to actually do about it
For a person: get on the judgment side of your own role. Deepen the understanding that lets you direct and check the machine rather than compete with it, because that is where value is concentrating, the same way the hard part of software was never the typing. For a business: stop asking whether AI will cut jobs and start asking which loops in your operation are pure grind. Those are the ones worth automating first, and doing it well tends to make the remaining roles better, not fewer.
The jobs question makes a good headline and a poor plan. The useful version is smaller and answerable: where is the loop, where is the judgment, and how do you move your people toward the second one. That is the conversation we actually have when we scope our solutions.
Prompted by videos including “The Future of IT Jobs in Next 5 Years” and reporting on whether AI creates more jobs than it cuts (2026). Credit to the sources; the framing is ours.
Saraswati Stitch®contact@saraswatistitch.com