Artificial Intelligence (AI) is changing the tasks through which young people have traditionally entered a profession or learned a field. New US labor market data suggests that workers at the beginning of their careers are more affected than experienced colleagues. At the same time, reports from schools, universities, and companies show that AI can save time without replacing subject knowledge, practice, or clear rules.
Why beginners face greater pressure
An updated Stanford study of the US labor market identifies a particularly strong change among workers aged 22 to 25. In occupations with high AI exposure, meaning jobs in which AI can perform or reshape many tasks, their employment was 19 percent below that of peers in less exposed fields. The gap had been 13 percent in the previous year, according to the study.
The researchers used anonymized, frequently updated payroll data from human resources company ADP. They assessed the AI exposure of occupations by combining a model of potential labor market effects with the Anthropic Economic Index, which examines how the Claude AI assistant is actually used across jobs. The data therefore shows an association, but it does not prove that AI alone caused every missing entry-level position.
The issue does not necessarily affect entire professions. It often concerns typical junior tasks such as organizing information, producing first drafts, summarizing material, or preparing basic analyses. Those assignments normally help beginners understand workflows and develop judgment, so automating them may raise short-term productivity while removing a stage at which new workers gain experience.
The hiring plans of OpenAI chief economist Ronnie Chatterji illustrate how fluid AI-related work can be even for specialists. He said job profiles in his research team of more than ten people keep changing and that employees need to be comfortable with discomfort. The team includes economists, data scientists, business specialists, and former government employees, suggesting that adaptability must be combined with substantial expertise.
Which skills turn saved time into useful time
Generative AI, meaning systems that create new text or other content in response to instructions, can substantially accelerate routine work. In User Experience Research (UX Research), the systematic study of how people experience products and services, it can transcribe interviews, sort answers, condense patterns, and draft initial user profiles. A practitioner article from a Swiss UX consultancy reports that some tasks that once took days can now be completed in hours.
Speed does not guarantee quality. If you ask weak questions, pursue unclear goals, or provide poorly structured data, AI merely produces a questionable result more quickly. Experience, empathy, and critical thinking remain necessary because AI mainly recombines existing information and cannot independently determine which problem deserves attention.
The first essential skill is defining the task. You should be able to specify the required result, decide which information may be used, and set criteria for a satisfactory answer. That is followed by prompting, the practice of giving an AI system clear instructions, as well as checking for errors, unsupported claims, and bias against the original material.
A practical household example comes from an analysis of browsing activity from more than 200,000 US households between 2021 and 2024. The observed users turned to ChatGPT for productive obligations such as tax returns, but then spent much of the saved time on platforms including TikTok and Instagram. A faster tool can create free time, but it does not decide how you use it.
Pros and Cons of AI in learning and work
Pros:
- Time savings – AI can produce transcripts, summaries, and initial drafts faster than a fully manual workflow.
- Low barrier to entry – You can request explanations, compare alternatives, and ask follow-up questions without waiting for a scheduled course.
- Less routine work – Repetitive preparation can be shortened, leaving more time for decisions, conversations, and careful review.
- Personalized practice – Learners can adjust explanations, request additional examples, and test their understanding over several rounds.
Cons:
- Lost practice – When AI handles junior tasks completely, beginners lose opportunities to build fundamentals and professional judgment.
- Unreliable output – Language models can state false or biased answers persuasively, making verification essential.
- Outsourced thinking – Copying an answer can complete an assignment without producing any understanding of the subject or process.
- Privacy risks – Confidential customer, school, or personal information may end up in unsuitable private AI accounts.
The boundary becomes particularly clear with AI agents, systems that can independently perform several digital steps instead of merely answering a question. According to a report on online courses at universities, some students instruct agents to open learning platforms, analyze course materials, and answer multiple-choice tests. Through connections to services such as Canvas, Brightspace, and Blackboard, the systems reportedly can even participate in chats with other students.
This goes beyond a new form of copying because the entire learning process, rather than only the final answer, is delegated. A passed quiz then reveals little about whether the student understands the underlying relationships. Schools and employers may therefore need to place more weight on a traceable work process than on a flawless-looking result.
What this means for your own practice
For beginners: Start with a limited task whose result you can verify yourself. For example, ask AI to summarize an existing text, compare the summary with the original sentence by sentence, and mark omissions or distorted emphasis. This teaches you both how to write precise instructions and how to recognize the system’s limits.
Then use AI as a learning partner rather than a substitute. You might ask for an explanation, solve the task yourself, and only afterward request criticism of your reasoning. Keep your own notes, sources, and intermediate steps visible so that you can reconstruct the thought process instead of merely possessing a finished answer.
For advanced users: Greater value comes from repeatable workflows with explicit review points. You could have several interview transcripts sorted into initial themes, but you still need to compare those themes with the original statements and search for counterexamples. Record the prompt, source material, changes made to the output, and the final human decision.
Advanced use also means recognizing when AI is inappropriate. Sensitive personnel records, customer data, and unpublished business information do not automatically belong in a publicly accessible chatbot. This is particularly relevant in Switzerland: a partner article from a provider in the St. Gallen Rhine Valley describes customer data being entered into private ChatGPT chats when companies provide tools without sufficient training or controlled workflows.
What schools and companies need to change
A license or a single course does not create AI competence. Swiss provider KI-Studio argues that some teams barely reopen a tool after training, while groups receiving continued support develop their own practical uses. This is a provider observation and has not been independently verified, but it supports a broader point: practical competence requires repeated use, feedback, and rules tied to real assignments.
Support remains uneven in education. In a Studyflix survey of 1,012 students, vocational learners, teachers, and parents, 74 percent said they had taught themselves how to use AI. Only four percent said a teacher had provided the relevant skills; because all respondents used the platform, the result cannot automatically be generalized to every learner.
Some educational institutions are responding with assessments that make independent thinking more visible. According to the report, students aged 16 to 19 in Denmark must orally defend major written assignments, while the University of Chicago Law School has banned laptops for first-year students during classes and exams. These measures are not necessarily intended to exclude AI completely, but to preserve independent and critical thought.
Another approach emphasizes general skills instead of a single product. Cheshire Academy in Connecticut does not require teachers to use AI and, according to MIT Technology Review, trained staff in prompting as well as the risks of incorrect and biased responses. Teachers use a mixture of general chatbots and specialized education tools while the school continues without one mandatory approach.
For Swiss schools and companies, the practical priority is not the number of accounts distributed. Learners and employees need to know which data they may enter, how to verify results, and when they must work without AI. This requires documented rules, assignments that reveal the reasoning process, and enough time to build subject knowledge; multilingual workplaces and classrooms must also check whether responses remain equally accurate and appropriate across the languages being used.
AI can accelerate entry-level tasks and make learning more personalized, but it can also displace the practice from which later expertise develops. Its most useful role combines time savings with independent verification, traceable workflows, and deliberate training. The unresolved risk is that organizations may measure immediate productivity while the long-term loss of junior experience becomes visible only much later.
Sources
- AI is hitting entry-level jobs hardest, Stanford study finds – Medium unknown, 2026-08-24
- OpenAI-Chefökonom sucht Mitarbeiter für einen Job, der sich ständig verändert – Medium unknown, 2026-08-23
- ChatGPT spart Zeit – die wir dann für Social Media verschwenden – Medium unknown, 2026-08-23
- KI im UX Research: starkes Werkzeug, kein Ersatz – Medium unknown, 2026-08-26
- KI-Lizenzen kaufen ist der einfache Teil. Das KI-Studio übernimmt den anderen. – Medium unknown, 2026-08-24
- „Wir lassen junge Menschen allein“: Warum Schüler sich den Umgang mit KI noch oft selbst beibringen müssen – Medium unknown, 2026-08-22
- Studierende lassen KI-Agenten komplette Onlinekurse absolvieren – so reagieren Lehrkräfte – Medium unknown, 2026-08-23
- How to encourage smarter AI use in the classroom – Medium unknown, 2026-08-24
