UBS just told its next class of junior bankers something no major global bank has said this bluntly before: know how to use AI, or don’t bother applying. Starting with its 2027 intake, the Swiss bank is folding AI job skills directly into the hiring bar for graduates and interns in its global banking and markets division — alongside academic transcripts and finance aptitude, not somewhere after them. It’s one bank’s policy change, but it points at something bigger: the skills that used to set ambitious candidates apart are turning into table stakes.

What UBS Actually Changed

According to the Financial Times, UBS is telling graduate and intern candidates that starting with the 2027 class, AI fluency will be judged alongside academics and finance aptitude as an explicit hiring criterion — not something the bank trains into new hires after they’ve already been picked. The bank wants applicants who can already use AI for research, modeling, and first drafts, and who know when the model’s output is wrong. UBS is reportedly the first major global investment bank to make AI literacy an entry gate rather than an on-the-job skill.

That last distinction is the real story. Plenty of employers have said “we use AI tools here” for a couple of years now. Almost none have said “if you can’t already do this, we don’t want to interview you.”

Why AI Job Skills Are Becoming a Baseline Requirement

UBS isn’t guessing that AI fluency matters — the data backing that bet is starting to show up elsewhere too. OpenAI recently reported that its own research organization now logs 3.1 agent-workdays of effort for every eight hours of human labor, up from a point in June 2026 where agent output still trailed human labor entirely. Daily inference use for the median researcher using coding agents rose to more than $600 at API prices by mid-August, with the top 10% of researchers running workflows costing over $7,000 a day.

That’s not a story about AI replacing researchers. It’s a story about what happens once someone genuinely knows how to direct AI at real work: their output multiplies instead of nudging upward. Employers who see that gap inside their own teams have every reason to screen for it before day one, rather than spend a year training it in — which lines up with a broader shift already visible in how fast enterprise AI agent adoption has grown in 2026. UBS is simply the first to say the quiet part out loud in a job posting.

The Specific Skills Employers Actually Mean

“AI proficiency” sounds vague until you look at exactly what UBS is asking for. Stripped of finance jargon, it breaks down into a skill set that applies well outside investment banking:

  • Using AI for real work product — first drafts, research summaries, or a first pass at a model, not just brainstorming ideas over lunch
  • Verifying before trusting — catching a plausible-sounding but wrong number or claim before it reaches a client, a boss, or a published document
  • Knowing a tool’s failure modes — where a specific model tends to be confidently wrong, and why, instead of treating every answer as equally reliable
  • Treating AI as leverage, not a substitute — using it to do more of your own judgment work faster, not to skip the judgment entirely

None of that requires a computer science degree. It requires actual hands-on time with the tools, applied to work that matters, with someone occasionally checking your homework.

How to Build AI Job Skills Before an Employer Asks For You to Prove It

The gap between “I use ChatGPT sometimes” and what UBS is describing is smaller than it looks, but it takes deliberate practice to close. A few concrete starting points:

Start applying AI to work you’d normally do unassisted — a research summary, a first-draft email, a rough outline — rather than only using it for casual questions. Our guide to mastering prompt engineering covers how to get a model to actually do useful work instead of generic filler, which is exactly the gap between using AI and being “proficient” in it.

Get in the habit of fact-checking what comes back. Pick a claim or number the model gives you and verify it against a real source before you’d ever use it for something that matters — that single habit is most of what “know when the model is wrong” actually means in practice.

If you’re still in school, this is the cheapest time to build the habit. Our piece on using ChatGPT to study smarter and our roundup of everyday ChatGPT use cases are both good places to start turning casual use into the kind of fluency an employer can actually see in an interview or a work sample.

What to Watch Next

UBS is the first major bank to put this in writing, but it’s unlikely to stay a finance-only story. Once one large employer in a competitive, credential-heavy industry makes AI fluency an explicit filter, competitors tend to notice fast — especially with data like OpenAI’s agent-productivity numbers giving them a concrete reason to. The practical takeaway isn’t “learn AI because a bank in Switzerland said so.” It’s that AI job skills are moving from a personal productivity hack to something you may genuinely need to demonstrate to get hired at all — and the people who started building that habit early won’t be scrambling when it shows up on a job listing.