Swiss bank UBS plans to make knowledge of Artificial Intelligence (AI), meaning software used for tasks such as text analysis and research, part of its hiring criteria. This brings into the workplace a debate already affecting schools and universities: you are expected to use AI without outsourcing your judgment or losing fundamental skills. Recent reports on applications, academic papers, and AI tutoring show how unsettled this new standard remains.
What does AI literacy mean at work?
Graduates and interns seeking 2027 roles in UBS Global Banking and Markets will be expected to show how AI can improve results and efficiency. According to the report on UBS’s new hiring requirements, candidates are set to face questions about their use of AI during interviews. This requirement will accompany a strong academic record and is also expected to apply to other newly advertised positions.
This means more than recognizing the name of a popular chatbot. Banks increasingly automate work traditionally assigned to junior employees, including financial analysis, research, and client presentations. UBS nevertheless says AI literacy should supplement academic and social abilities rather than replace them. Santander is also looking for advanced AI users in some of its trainee programs, according to documents cited in the report.
Candidates therefore face two connected demands. You need to use a tool productively while judging whether its output is professionally sound. Someone who can produce an analysis faster but does not understand its foundations has mastered only the more convenient half of the task. That concern matches a warning from JPMorgan’s European head that junior employees must not lose their fundamental skills.
The relevance to Switzerland is direct: UBS is a major Swiss company explicitly adding AI ability to personnel selection. The report does not show that the same condition already applies throughout the Swiss labor market. It instead illustrates changing expectations in a field where research, analysis, and presentations are part of everyday work.
How is AI changing applications and academic papers?
AI can help candidates revise resumes and cover letters. Employers use the technology as well: a Randstad survey found that 33 percent of companies using AI in recruiting apply it to the analysis of application documents, while another 31 percent use it to support candidate preselection. AI literacy is therefore relevant on both sides of the hiring desk.
Some applicants are attempting to influence automated reviews by placing hidden instructions in their resumes. This technique is known as prompt injection, meaning commands embedded in content that will be processed by an AI system. A report on hidden resume instructions describes a case in which screening software flagged an anomaly in an application for a legal and compliance position.
A difficult entry-level employment market provides context for these tactics. An analysis of more than four million job postings from January 2020 through April 2025 found that the share of advertised entry-level positions had fallen significantly. In an accompanying survey, graduates under 30 reported submitting an average of 40 applications to receive one interview invitation and spending about seven hours on each application. That burden explains the demand for automation, but it does not justify manipulation.
At universities, choosing an AI tool has also become part of the work. In an August 2026 blind test based on 6,851 anonymous votes, students preferred Google Gemini for writing tasks in 39.6 percent of cases. Claude received 31.8 percent, while ChatGPT received 29.2 percent. The comparison of Gemini, Claude, and ChatGPT, however, measured which responses participants liked best, not whether an entire academic paper was factually correct or complied with university rules.
The practical lesson is less definitive than the ranking might suggest. A model may produce a clearly structured essay draft, but you remain responsible for the arguments, evidence, and your own voice. The test suggests that students preferred balanced responses that appeared complete, organized, and clear. It does not establish that the leading tool performs best in every subject or on every assignment.
What do schools reveal about useful limits?
New York City is drawing a much stricter line for younger students. About 600,000 children in public schools will not be allowed to use AI tools through eighth grade starting with the new school year. Individual screens are also prohibited through third grade, while companion chatbots, digital systems designed to act as conversational companions, are banned at all grade levels.
Teachers may continue using AI to prepare lessons, but not to grade students. For middle school students, the education department recommends no more than 45 minutes of screen time per day. A task force made up of teachers, politicians, experts, and union representatives is expected to produce a report by April 2027 to guide future school policy.
The New York restrictions for public schools reflect more than general skepticism about technology. They address concerns that children could hand logical reasoning over to software before developing it for themselves. Some parent groups want to go further and are calling for a complete two-year moratorium on AI in classrooms.
A two-year study across 18 middle schools in Tennessee offers a more nuanced picture. Selected students received access to Khanmigo, an AI tutor from Khan Academy that was designed to guide them through math problems rather than provide answers directly. Math performance rose by 1.3 percentile ranks, positions within a comparison group, per semester; over a full school year, the gain was 0.06 to 0.08 standard deviations, a statistical measure of distance from the average.
The effect reached 0.14 standard deviations among students who used the tutor consistently. According to the researchers, however, those gains were similar to improvements achieved through ordinary practice with Khan Academy without AI. Although 96 percent tried Khanmigo at least once, many rarely used it for learning, sent unrelated messages, or attempted to obtain direct answers. The report on the long-term AI tutoring study therefore supports human supervision rather than simply providing access to a tool.
Pros and Cons of AI as a Core Skill
Pros:
- Workplace relevance – Employers such as UBS are testing AI skills as analysis, research, and presentation work becomes partly automated.
- Practical assistance – AI can help revise application materials, structure writing tasks, and explain learning material repeatedly.
- More efficient preparation – New York teachers may still use AI for lesson planning even though younger students face strict limits.
- Informed tool selection – The student blind test shows that different models can be perceived differently on the same writing task.
Cons:
- Loss of fundamentals – If you fully outsource analysis, reasoning, or writing, you may weaken or fail to develop essential academic and professional abilities.
- Limited learning gains – The AI tutor produced only small improvements that researchers said were also achievable through practice without AI.
- Risk of manipulation – Hidden commands in application documents can influence automated selection systems and undermine fair hiring.
- Unclear assessment standards – A popular or polished answer is not automatically correct, independently produced, or permitted under university rules.
What does this mean for your own practice?
If you are a beginner, start with a narrowly defined task. You might ask AI to review a structure you wrote yourself or explain a math solution step by step instead of requesting a finished application or academic paper. Then compare the response with your original text, course material, or your own calculation.
For advanced users, additional value comes less from longer prompts than from systematic comparisons. You can process the same assignment with different instructions or tools, mark the differences, and explain which answer fits the subject and writing context better. The university blind test demonstrates why a familiar brand name is not a substitute for this evaluation.
In an interview, explaining a concrete workflow is also likely to be more meaningful than listing several product names. The UBS example indicates that the focus is on better results and efficiency while academic and social abilities remain necessary. One transparent example would be using AI to prepare research, then checking its claims yourself and selecting the relevant material for a presentation.
In schools and training programs, competence also includes knowing when not to use a tool. The New York rules and the Khanmigo study take different approaches, but they point to the same educational problem: access alone does not create discipline or understanding. The sources provide no comparable nationwide school policy for Switzerland, so the New York model cannot simply be treated as a Swiss template.
AI literacy is moving from optional knowledge toward an expected ability in work and education, but it remains incomplete without subject expertise, judgment, and transparent rules. The examples range from useful research support to attempts to manipulate automated hiring systems. The unresolved risk is that institutions may demand AI skills faster than they establish fair assessment standards and effective educational boundaries.
Sources
- Schweizer Grossbank UBS macht KI-Kompetenz zur Einstellungsbedingung – Unknown, 2026-09-07
- New York City verbannt KI-Tools aus öffentlichen Schulen bis zur achten Klasse – Unknown, 2026-09-07
- Langzeitstudie zeigt: KI-Nachhilfe funktioniert nur mit menschlicher Betreuung – Unknown, 2026-09-06
- Beste KI fürs Studium: Warum Gemini bei Hausarbeiten besser als Claude oder ChatGPT ist – Unknown, 2026-09-05
- Versteckte Befehle im Lebenslauf: Wie Bewerber per Prompt-Injection aus der Masse herausstechen wollen – Unknown, 2026-09-06
Image: Mikhail Nilov via Pexels