My son just finished his first year of college. A few months ago we were sitting at dinner and he asked me something I was not expecting. He said: "Dad, is it even worth picking a career right now? Like, how do you know what's going to exist?"
I did not have a great answer. I gave him the responsible parent version of one, something about adaptability and fundamentals and how the best careers have always been built on skills that travel. He seemed reasonably satisfied. I was less satisfied with myself.
Because the honest answer is that I think he is asking exactly the right question. And I am not sure the adults who should know have fully worked it out yet.
That conversation keeps coming back to me as I spend time in the job market, reading descriptions, talking to recruiters, watching companies try to figure out AI strategy in real time. One pattern keeps surfacing, and it does not get talked about enough.
I do not think the first major workforce impact of AI is going to be experienced workers losing jobs.
I think it might be beginners losing the opportunity to become experienced in the first place.
For decades, companies carried a layer of junior work that was never really about efficiency. It was about development. New analysts sat in meetings they barely understood. New coordinators rewrote things five times. New developers pushed code that broke things and learned why. The first draft, the meeting summary, the research synthesis, the repetitive coordination, the documentation everyone avoided. All of it was tolerated because the organization understood something important: repetition is how people build judgment.
That layer is now the exact layer AI is best at compressing.
This year at Google I/O, the company made clear it sees the future as increasingly agentic. Not just AI answering questions, but AI coordinating workflows, generating deliverables, performing research, and operating more independently inside day to day work. Around the same time, Anthropic leadership talked openly about how AI now generates most of the company's code, shifting engineers toward oversight and review rather than building components themselves.
That is not theoretical. That is already operational. And buried inside that shift is something nobody seems quite ready to say out loud: a lot of entry level work lived exactly in those tasks.
Companies increasingly expect polish immediately. You can feel it in hiring conversations. You can feel it in job descriptions asking entry level candidates for strategic thinking, executive communication, AI fluency, and years of experience all at once.
The economics of patience are changing. Once a company realizes one experienced person with strong AI tooling can absorb work that previously required two or three coordinators or analysts or support layers, the pressure to flatten organizations becomes very real. The productivity gains are genuine. I use AI constantly myself. But two things can be true at once: AI can improve productivity while quietly weakening the development pipeline that used to turn beginners into experts.
The NFL has never once drafted a rookie quarterback and expected them to immediately read defenses like a ten year veteran. The entire system is built around development because everyone understands that talent and experience are not the same thing. Film sessions matter. Practice reps matter. Sometimes sitting behind a veteran for two seasons matters. The reps are not inefficiency. The reps are the point.
Work used to operate with more of that understanding, even if imperfectly. There was at least some acceptance that becoming good at something required time, observation, and gradual exposure to complexity. The awkward middle stage mattered because it was where real learning happened, not from polished output but from the process of producing it badly first.
I had coffee recently with someone early in their career who said something that stuck. They told me they feel pressure to use AI for almost everything because everyone else already does, but they also worry they are skipping parts of the learning process without realizing it. That is a very modern fear. Not fear of the technology itself. Fear of becoming dependent on acceleration before building deeper understanding.
I do not think we fully understand the long term implications of that yet. Especially in leadership environments. Especially in moments where systems fail, priorities conflict, or organizations need real judgment rather than polished output. Because eventually every company still needs people who understand how the work actually works underneath the surface. And historically that expertise came from years of operational repetition that nobody thought twice about tolerating.
The expectation curve shifted fast, and younger professionals can feel it even if they cannot fully articulate it. You see it in job descriptions that ask for strategic thinking and executive presence and cross functional influence and five years of experience for roles that pay like internships. Sometimes it really does feel like companies want people ready for the playoffs before they have played a regular season game.
Maybe AI will ultimately create entirely new categories of work and development. History suggests technological shifts usually do, and I suspect that will happen here too. But that transition takes time, and right now we are in the part of the story where the old development model is eroding faster than the new one is forming.
My son asked me if it is even worth picking a career right now. The real answer is probably: yes, but the path to becoming genuinely good at something is going to require more intentional effort than it used to, because the organic repetition that used to build experience is no longer guaranteed.
That feels like one of the most important workforce questions of the AI era. Not whether humans get replaced. Whether humans still get enough opportunities to slowly become excellent in the first place.
My son's generation is growing up in a moment where the traditional signals of "this is a safe path" no longer apply the way they used to. The majors that seemed reliable are uncertain. The entry points that used to exist are compressing. And they can sense it even when adults are not saying it directly.
What I keep coming back to is that the most durable thing I can help him build is not a specific skill set. It is a tolerance for ambiguity and the ability to keep learning when the ground keeps moving. That might be the actual curriculum for this moment, whether or not anyone has figured out how to teach it yet.
Job descriptions have gotten longer and more contradictory. Entry level postings routinely ask for things that used to be mid to senior requirements. The justification, when companies offer one, is usually some version of "we need people who hit the ground running." Which is understandable. And which also quietly eliminates the conditions under which hitting the ground running becomes possible in the first place.
I am also seeing more experienced professionals in the job market than I would have expected. Not because they were displaced directly by AI, but because companies made headcount decisions, consolidated roles, and leaned on existing staff with AI tooling. The ladder got shorter from both ends simultaneously.
At the same time companies are eliminating junior roles, a lot of them are quietly struggling with knowledge transfer and institutional memory. The experienced people who used to mentor the junior layer are now fully occupied running larger workloads themselves. Nobody has figured out the long term consequences of that gap yet. But give it five years.