Jobs disappearing versus jobs created: the only number that is actually yours

Every forecast about AI and work reports the same shape of number. The World Economic Forum projects 92 million jobs displaced and 170 million created by 2030, a net gain of 78 million. It is a real figure and it is useless to you, because you do not hold 78 million jobs. You hold one, and the only question that matters at your scale is whether the one you hold is on the shrinking side, and if it is, which growing job your skills already reach.

We can answer part of that from our own data, and we have to be honest about the part we cannot. Our corpus is a snapshot of who is being hired now, not a time-lapse of who will be automated later, so it does not see jobs disappearing. For that, the government projections are the source. What our data does see, better than any projection, is the bridge: where one occupation’s skills already reach another.

The created jobs are real, but not the famous one

Start with what AI made. The new roles are in our corpus now, and they sort into two piles: the builders, well-paid and growing, and the support-and-hype layer, thin and already fading.

The split is the finding. The builders, AI engineer and machine-learning engineer and computer vision engineer, are real software-engineering jobs with an AI specialty, and they pay for it: computer vision clears 149,000 dollars. The bottom two are the ones the headlines named. Prompt engineer and data annotator are both low-demand, and the annotator, the human who labels the data that trains the models, sits at 63,000 dollars, the price the market puts on feeding AI rather than building it.

The lesson in that contrast is worth more than the totals. Prompt engineer was the job every 2023 headline named as the face of AI work. Two years on it is 74 postings in our corpus and fading, its tasks folded back into the broader AI-engineer role that actually grew. The created jobs are real; the ones named first are usually wrong. Betting a pivot on the meme job is how you arrive a year late to a role that no longer exists.

The disappearing jobs, and the bridge out of them

For the shrinking side we defer to the BLS (US Bureau of Labor Statistics) Employment Projections, which have named the decliners for years: cashiers, data-entry keyers, telemarketers, word processors, the routine roles automation reaches first. Our corpus mostly cannot see them, because the roles it carries are the ones still hiring. But for the occupations everyone calls automation-exposed, our adjacency graph shows something the decline projections never do: where the skills already go.

These are not consolation prizes. A medical assistant, in one of the fastest-churning jobs in the country, already covers 76 percent of what a nurse-practitioner posting asks for, the exact ladder that turns an automation-exposed role into one of the stickiest careers there is. An IT-support worker covers 54 percent of a network engineer; a bookkeeper, 55 percent of a financial controller. The move is up and sideways at once, and it is measurable today, before any decline forces it. Notice the shape: the strongest escapes stay inside the same world, healthcare into healthcare, tech support into tech, because that is where the skills already overlap.

Which skills carry you across

The bridge is not luck; it is a specific set of skills that appear on both sides of the shrink-to-grow gap. When we counted the skills that show up in the most different occupations, the winners were not any field’s headline tools. They were the portable ones: project coordination, data analysis, writing, and the handling of people under pressure, the competencies that travel because no single job owns them. A bookkeeper reaches a financial controller on ledger fluency and process discipline; a customer-support specialist reaches an executive assistant on judgment and scheduling. The skill that automates is the narrow, repeatable one; the skill that carries you is the general one, which is the same reason it was never the thing AI came for first. The full ranking is in our piece on the most transferable skills of 2026.

Why the net number is a trap

The 78-million-net figure hides the only thing an individual needs to know. Net creation can be strongly positive while your specific occupation halves, because the created jobs and the destroyed ones are different jobs, held by different people, often in different places. The macro number reassures the economy and abandons the worker. The micro number, the coverage between where you are and where the hiring is, does the opposite. It ignores the economy and tells you your next move.

That is the entire design of the instrument: it does not forecast whether AI will take your job, a question no one can answer honestly. It measures which growing jobs your current skills already reach, which our data can answer for any starting point. The new job titles are one half of the picture and the gravity wells are the other. Run your own. The net number is not yours. The bridge is.