The office roster looks much like last year's. The desk layout and the team names are the same. The one thing that has changed is that the two or three new hires who arrived on that floor every spring are not there this year. Changes that statistics do not capture usually come in this shape.
Warnings about AI and employment came one after another from international organizations and the Davos stage, and in Korea a related bill was introduced. A major American daily closed its sports section, and in the legal field the hiring barrier for those who have just been licensed has risen noticeably. Each looks like a separate story, but they meet at one point. The share that shrinks fastest in an organization is the work assigned to first-year employees.
Here the AI employment shock is not captured by an aggregate indicator such as "a few percent drop in total jobs." Policy should place its priority on redesigning the path by which people enter an occupation rather than on binding the job security of existing staff by law.
The reason is simple. The work automation handles well right now is repetitive work with a narrow range of correct answers, such as first-pass case-law searches, writing up game results, summarizing meeting minutes and drafting basic code. But these tasks also happened to be the training process through which new hires spent three years getting the work into their hands and building a feel for the job. Once the training ground disappears, neither the middle managers of five years from now nor the experts of ten years from now get made.
The counterargument is strong. It is that a vanished entrance has always been replaced by a new entrance. After typists and telephone operators disappeared, jobs that had not existed before appeared, and the labor market built a new step each time. The point that companies will cut hiring itself if headcount is locked in by regulation, which leaves new hires worse off, is also valid.
Still, what makes this time different lies in the order and the speed with which the technology spreads. Past technologies mostly spread in a way that added to the lower end of skill, but today's tools absorb junior work first and move upward. Even if new jobs appear, if those jobs also open in a form that requires three or more years of experience, no door opens for someone starting work for the first time. The total volume of replacement may add up, but the distribution across generations comes out wrong.
What is needed, then, is a mechanism that rebuilds the entrance rather than a device that prevents dismissals. Things such as an indicator that counts how many entry-level hires a company kept, a system in which the cost is shared when a company trains first-year workers, and public projects that let people build an actual work record in place of certificates. If the bill that was introduced is refined with a focus only on protecting total employment, the floor that is actually emptying will remain as it is.
The point to watch is the wording of this year's second-half open recruitment notices. How many listings are marked "experience not required," and whether those listings actually have a supervisor assigned to them. In place of the broad tally of hiring volume, those two lines show the labor market of the next five years more accurately. Whether organizations keep a new hire's right to make mistakes in the budget, or hand it over to the tools and erase it, will divide people's next decade.
