Who Builds the Next Senior?

The question at the center of the changing entry-level labor market is now precise: who builds the next senior when generative AI absorbs the work through which beginners once learned? The evidence does not support a simple answer that AI either destroys careers or upgrades everyone. It shows a more conditional pattern—AI can raise productivity and shift jobs toward expert responsibility, while the institutions that develop judgment can become weaker at the same time. The practical task is to judge whether an AI-redesigned role is building future expertise or merely demanding it.

The answer is not that routine work must be preserved unchanged, nor that beginners should avoid AI until they have somehow become experienced. The answer is that schools, employers, and hiring systems must deliberately allocate authentic practice, feedback, verification, and increasing responsibility. They must treat those conditions as part of the infrastructure of expertise rather than as leftover time that can be removed whenever a faster production path appears.

Growth without a guaranteed pathway

The broad labor-market evidence and the software-engineering case operate at different scales, but they illuminate the same division. PwC’s 2026 Global AI Jobs Barometer finds that AI-exposed firms in its measured sample experienced strong productivity growth. In 2025, the most-exposed quartile recorded 33.5% growth in turnover per employee from a 2018 baseline, compared with 24.0% for the least-exposed quartile. The same comparison showed faster headcount and wage growth among the most-exposed firms. AI exposure, in other words, can be associated with organizational expansion rather than labor substitution alone.

The report also describes a two-track occupational market. In professionalised work, AI takes on more basic tasks while people retain more expert responsibility. In democratised work, AI takes on more complex work and people remain with less expert tasks. Professionalised postings grew faster than democratised postings in the report’s comparison, and their advertised salaries and demanded skills grew faster as well. Newly added tasks in highly exposed occupations were more human-intensive on the EPOCH measure, relying more on Empathy, Presence, Opinion, Creativity, and Hope.

Those findings make the opportunity case plausible, but they do not settle the entry-level question. A professionalised role can grow while becoming more selective. An employer may want people who can interpret evidence, manage stakeholders, make strategic decisions, and take responsibility for consequences. The posting can show that those capabilities are in demand; it cannot show that the organization gives a beginner a safe way to practice them. The higher-value destination and the route into it are different things.

That difference appears in the early-career data. PwC finds that postings requiring zero to two years of experience in the highest AI-exposure quartile have flatlined relative to 2012, while lower-exposure groups continued to grow. At the same time, the most-exposed entry-level roles were more likely to demand traditionally senior capabilities, and seniorised roles within the top exposure quartile grew while non-seniorised roles declined. The labor market is not simply eliminating every junior opportunity. It is narrowing some conventional entry points and raising the expectations attached to the roles that remain.

The quantitative evidence must be held at its proper strength. The early-career trend is based on a limited country sample, dominated by U.S. postings, and uses years of experience as a proxy. PwC cautions that exposed entry-level conditions began weakening before ChatGPT’s public release and that economic and occupational changes may contribute. Its firm results come from larger formal surviving companies with usable data, not the whole economy. The professionalised and democratised categories are occupational mappings, not records of individual promotion.

These limits do not make the pattern irrelevant. They define the judgment it can support. The evidence is strong enough to say that work is being reorganized unevenly, that expert responsibility is becoming more valuable in some exposed roles, and that entry-level access is not moving in lockstep with productivity. It is not strong enough to say that GenAI alone caused the bottleneck, that every professionalised job offers mobility, or that every worker should pursue the same path.

The interviews by Sumin Yu and Taesup Moon explain how the missing link may form. Their study involved 14 software engineers in South Korea: eight juniors at the threshold of entering the workforce and six seniors with at least six years of experience. The sample is small and purposive, and the study is cross-sectional. Its value is not prevalence. It is the mechanism that the labor-market aggregates cannot show directly.

The researchers call that mechanism Absorption. A senior engineer with accumulated context uses generative AI to handle routine code generation, initial debugging, or documentation directly. Work that once reached a junior no longer needs to be delegated.