Will AI Really Replace the Careers My Child Is Preparing For Today?

I code with AI all day long. Not as a spectator, as a practitioner. I know what these tools do well, what they do badly, and what they probably will never do. So when a parent reaches out to ask me “should my child still study law, given AI?”, I can answer with something other than newspaper headlines.

This article is an honest assessment, profession by profession, of what generative AI actually changes in 2026, and of what that means for a student choosing a path today. No doom-mongering, no starry-eyed techno-optimism.

What AI Does Well in 2026 (and What the Media Exaggerates)

Before we talk about careers, we need to understand what generative AI can actually do today:

• Drafting a first version of a text (a contract, an email, a report, an article) at a decent but generic level
• Summarizing a long document in 30 seconds
• Translating at a quality often superior to a non-native speaker
• Coding standard features from a plain-language description
• Analyzing structured data and pulling patterns out of it
• Conversing fluently and answering factual questions

What it cannot do (and the list is long):

• Judgment: in the ethical, contextual, nuanced sense. A judge, a doctor, a leader exercise a form of judgment that AI cannot reproduce.
• Innovation: it recombines what already exists, it does not create something fundamentally new.
• Handling relational uncertainty: negotiating, persuading, supporting a patient in distress, defusing a team crisis.
• Being accountable: legally and morally. When a lawyer signs a contract, they put their own liability on the line. AI cannot be held accountable.
• Reading implicit context: it systematically misses the unspoken, the political stakes, the power dynamics.

Profession by Profession: What Really Changes

Lawyer / legal professional

What AI already does: case-law research, drafting first versions of standard contracts, summarizing long files, analyzing clauses, documentary due diligence.

What it does not do: argue a case in court, negotiate with the opposing party, advise a client in crisis, interpret an ambiguous law in a novel context, take responsibility for a legal opinion.

Verdict: the “pure research” junior lawyer role is going to change: less time spent searching, more time spent analyzing and advising. But the need for lawyers is not going to fall. It is going to shift: more high-value advisory work, less mass documentary production. For a student who loves the law, the answer is: go for it, but also learn to use AI as a tool.

Doctor

What AI already does: diagnostic support on medical imaging (radiology, dermatology), analysis of biological data, drafting reports, triaging emergencies.

What it does not do: examine a patient, listen for the symptoms they do not say out loud, make a treatment decision under uncertainty, deliver a serious diagnosis, operate.

Verdict: AI is a medical co-pilot, not a replacement. Doctors who know how to use it will be more effective than those who do not. But human contact, clinical judgment and accountability remain irreplaceable. For a student who wants to go into medicine: the profession will evolve, not disappear.

Engineer

What AI already does: code generation, design optimization, simulation, analysis of industrial data, predictive maintenance.

What it does not do: design a complex system from A to Z, arbitrate between conflicting constraints (cost, safety, deadline, environment), lead a project involving humans, take responsibility for a structure or a piece of software.

Verdict: tomorrow’s engineers will use AI heavily as an accelerator. But engineering has always been a profession of technical judgment and managing uncertainty, and that does not change. For a student who loves technical work and problem-solving: this may be the profession that benefits the most from AI, not the one that suffers from it.

Accountant / auditor

What AI already does: bookkeeping and reconciliation, anomaly detection, report production, compliance analysis.

What it does not do: interpret accounting standards in complex cases (M&A, international consolidation), advise a leader on tax optimization, manage the client relationship.

Verdict: this is the profession most transformed by AI in the short term. Repetitive tasks (data entry, reconciliation, producing financial statements) are largely automatable. But advisory expertise and normative judgment remain human. For a student: “pure execution” accounting is not a good bet. “Advisory plus expertise” accounting is still solid.

Teacher

What AI already does: personalized tutoring in mathematics, languages, programming. Automatic grading. Generating exercises tailored to the student’s level.

What it does not do: motivate a disengaged student, manage a classroom of 30 teenagers, pass on the love of learning, spot distress, teach critical thinking.

Verdict: the teacher’s role is going to refocus on what AI cannot do: human support, socialization, teaching judgment. The “pure knowledge transmission” tasks will be increasingly AI-assisted. For a student who wants to teach: the profession has a clear future, but it will change in nature.

Developer / IT professional

What AI already does: write standard code, debug, refactor, write tests, document.

What it does not do: understand a client’s business need, architect a complex system, arbitrate technical trade-offs, manage technical debt, work as a team on a long-term product.

Verdict: paradoxically, this is the most exposed AND the most resilient profession. In the short term, the junior market has contracted: developer job postings have fallen sharply since the 2022 peak, with a marked decline among 22 to 25-year-olds according to a 2025 Stanford study. Over the longer term, the projections stay positive: the US BLS forecasts roughly +15% more developers by 2034. For a student, the lesson is clear: aim for versatility, not code alone. Knowing how to code with AI remains a force multiplier, provided you also understand the field you are building tools for.

The Three Real Questions to Ask

Rather than “will AI replace job X?”, the right questions are:

1. Does the job rest mainly on human judgment or on repetitive execution?

If it is judgment (doctor, advisory lawyer, architect, manager, researcher): the job transforms, it does not disappear. If it is repetitive execution (bookkeeping data entry, simple translation, mass documentary production): real danger.

2. Does the job involve irreplaceable human contact?

A psychologist, a salesperson, a teacher, a caregiver rely on human contact in a way that AI does not reproduce. These are jobs that technology can assist but not replace.

3. Is my child ready to keep learning continuously?

This may be the only question that really matters. Every profession is going to evolve, some quickly, some slowly. The student who will be best equipped is the one who knows how to learn new things, not the one who has memorized a fixed syllabus.

What I Tell the Families Who Reach Out to Me

When a parent asks me “my child wants to do X, is that still viable?”, my answer is almost always the same:

Yes, it is viable, as long as the child chooses that path because it genuinely interests them, not because it “pays well” or because it is “safe.” AI is going to upend the hierarchy of “safe” jobs: the careers considered “gold-plated” today (quantitative finance, auditing, business law) are more exposed to disruption than jobs considered “modest” (caregiver, craftsperson, educator).

The best investment a student can make is to choose a path they are passionate about AND to build, in parallel, a solid digital literacy, not necessarily coding, but understanding how AI works, what it can do, and how to use it in their own field.

Key Takeaways

• AI does not replace jobs, it replaces tasks. Most jobs are a mix of automatable tasks and irreplaceable human judgment.
• The most exposed jobs are the ones that rest on repetitive execution (data entry, documentary production, simple translation). Jobs built on judgment, human contact and creativity stay solid.
• For a student, the question is not “which job is safe from AI” but “am I ready to keep learning continuously and to use AI as a tool.”
• Choosing your path out of passion rather than by calculating for safety has never been more relevant, because safety calculations based on the old world are no longer valid.

Going Further

• Anthropic: Claude (our own daily working tool)
• OECD: AI and the Future of Work
• France Stratégie: Impact of AI on Employment

---

Article written by Constantin Mardoukhaev, co-founder of Axiom Academic. Constantin codes with AI every day and supports French-speaking families in their orientation choices, including when those choices run up against the question of the future of work.