The job can move before you get through the door
You studied finance. You worked on your CV. You practised explaining why you want to be a banker. Now there is another thing to prove.
Can you use AI to improve the work and still take responsibility for the answer?
That is a tougher requirement than adding an AI tool to a skills list. It asks for evidence. For someone trying to win a first banking job, the pressure arrives before there is a salary, a manager or a training budget to help pay for the preparation.
What UBS is asking applicants to show
The Financial Times reported on September 6 that AI proficiency is becoming a recruitment requirement for junior bankers entering in 2027, with questions being added to graduate and internship assessments.
UBS’s own Global Markets graduate posting for London asks for evidence that responsible experimentation with AI can improve efficiency and results. The Global Banking graduate posting for Hong Kong contains the same expectation.
These are examples from named roles and locations. Candidates should read their exact vacancy. An investment banking graduate requirement should not be casually extended to every job, country and division at UBS.
This is not another UBS layoff announcement
The new policy changes what applicants are expected to demonstrate. It does not identify a group of junior employees being dismissed or a number of graduate seats being removed.
Our UBS Credit Suisse integration investigation covers the separate restructuring and workforce reduction story. A bank can consolidate existing operations while continuing to recruit for other roles.
The connection worth watching is whether AI expectations rise while comparable intake numbers fall. Until there is evidence of both, calling this policy an AI layoff would tell a more dramatic story than the facts support.
Knowing the tool is not the same as doing the job
A polished answer can still contain a wrong number, an invented source or a conclusion that the evidence does not support. In banking, speed without checking can make a weak piece of work more dangerous.
A useful candidate can explain the task, the source material, the steps the tool helped with and the decisions that stayed human. They can also describe where the output failed.
If your only example is that AI wrote a confident paragraph, you have shown that the software can produce prose. Show what you did to make the result useful.
The work juniors learn on is part of the change
The Hong Kong Global Banking description includes company evaluation, financial modelling, valuation, industry analysis and materials for pitches. The London Global Markets role includes interpreting market information and supporting client activity.
Those duties contain different kinds of work. Gathering public information and preparing a first draft are different from choosing sound assumptions or deciding what a client should be told.
Our assessment is that applicants will need to show where AI helps within a task, rather than talk vaguely about replacing a whole profession. The useful unit is the work you can explain and check.
The training has not disappeared
UBS’s Graduate Talent Program page describes an AI Fluency Pathway alongside its broader training. It focuses on practical applications, responsible use and sound judgement.
That matters because the story is not that graduates must arrive knowing everything. The bank is still describing development after entry. The sharper change is that some evidence of useful AI practice is now part of getting considered.
Applicants should ask what they are expected to know on arrival and what the programme teaches. Clear entry standards are useful. A vague demand to be an expert in everything is not a training plan.
Build one example you can defend
For practice, choose a public annual report and a small question, such as how one business division’s revenue changed. Read the relevant pages yourself before using an AI tool to organise the material.
Then verify every figure against the filing. Check the reporting period, currency and whether the comparison uses the same business scope. If the tool invents a reason for the change, remove it unless the source supports it.
Explain what the tool contributed, what you corrected and how the final answer became better. If you measured time saved, include the checking time too. An honest modest improvement is more credible than a spectacular claim you cannot reproduce.
Your judgement is the part worth showing
The practice example above is our suggestion, not a disclosed UBS test. It works as preparation because it lets you discuss a real task without pretending to know the interview questions.
Be ready to explain when you would avoid AI. A task may involve confidential information, an unsuitable tool or a decision that needs an authorised reviewer. Knowing that boundary belongs in the answer.
Use public or synthetic material for personal practice. Follow the employer’s instructions during any assessment, including restrictions on AI assistance. Being able to work with AI is not permission to let it answer a test that requires your own unaided work.
Do not buy panic with a certificate
A new hiring requirement can make every advert for an AI course feel urgent. Before spending money, identify the gap you actually need to close. Is it understanding financial statements, checking output or explaining a useful workflow?
The postings reviewed do not name a particular paid AI certificate as the answer. Completing a course may help you learn, but the badge alone does not demonstrate how you work.
Start with a task you can finish, check and explain. Keep applying while you improve. Waiting until you feel perfectly qualified can become another way of missing the recruitment window.
Candidates should not have to guess the rules
Employers gain when applicants arrive with useful skills. They also have a responsibility to make the assessment understandable. Candidates need to know what tools are permitted, what will be judged and what support is available.
Ask the recruiter whether the assessment measures practical use, judgement, technical depth or all three. If a tool or format creates an accessibility barrier, raise the need for an adjustment through the recruitment support process.
A candidate with strong judgement and limited access to expensive software should have a fair way to demonstrate ability. Familiarity with a paid product is a weak substitute for a clear assessment of the work.
Already inside the bank? Watch the new baseline.
An entry requirement can become a benchmark existing employees feel pushed to match. That is a possible next step, not a confirmed UBS policy for every current banker.
Ask what training time is available, which tools are approved and how useful outcomes will be measured. A higher target should account for review, corrections and the parts of the job that still take human attention.
Our AI productivity trap investigation follows the wider problem of saved time becoming more work. This UBS story concerns recruitment first. Existing workers should watch whether the expectation moves into their own goals.
Count seats before declaring the junior job dead
The next meaningful evidence would be comparable graduate class sizes, internship places and offers. Compare the same business, location and recruitment stage. A global applications figure cannot tell you how many seats exist on a London desk.
Fewer seats would still need an explanation. Deal activity, budgets, business priorities and integration decisions can affect recruitment alongside AI. A change in headcount is not enough to identify its cause.
Follow conversion from internship to a permanent offer where the figures are available. Hiring headlines say a door is open. Actual offers show how many people get through it.
Do not build your entire future around one logo
Prepare seriously for UBS if the role fits. Also compare vacancies at other banks and related employers by the work, training and entry requirements they offer.
An application can be strong and still fail in a competitive process. That does not prove you are obsolete. It means you need more than one route into paid work and more than one employer judging your potential.
Before accepting an offer, use our guide to checking an employer and the role’s funding. Winning a job is only part of the decision. Understanding what you are joining matters too.
Quiet Power: make the skill useful to you too
Keep a short, nonconfidential example of a task you improved. Explain the original problem, the checks and the final result in your own words. That evidence should travel with your career rather than depend on one employer’s internal tool.
For employees, ask: “Which parts of my role should improve with AI, what training time is available and how will the review work be reflected in my targets?”
If you are asked to document or transfer substantial parts of your work, our guide to training an AI replacement covers that separate situation. A learning request and a role transfer deserve different questions.
Three free tools to follow the pressure
The Layoff Tracker + Corporate Stress Index follows 50 technology and banking employers, including UBS. Check its supporting sources and weekly history when researching the employer behind a vacancy. A company pressure score is not a measure of the safety of an individual graduate seat, and this recruitment requirement should not be counted as a completed layoff.
Get the free Weekly Layoff Intelligence Report to follow hiring changes, AI pressure and restructuring beyond one application cycle. It gives candidates and employees a regular view of what employers are changing while they keep building their own options.
Already working and seeing AI expectations arrive beside missing backfills or changing duties? The free Job Threat Check asks seven questions about pressure around your company, team and role, including lost influence and transferred work. You receive a score, explanation and practical next steps before the optional email invitation. It assesses workplace pressure; it does not grade AI skills or predict whether UBS will hire you.
The Grind Hotline Read
The first employment consequence of AI may be a higher bar for a job that still exists. Workers need to recognise that change without being told their entire career has already disappeared.
Learn the tools. Learn how to catch their mistakes. Keep enough command of the work to explain an answer without the machine beside you. A bank can buy software. Your case for being hired needs to show what you add.
Sources and verification
Checked September 6, 2026 against the FT recruitment report, UBS’s London Global Markets posting, its Hong Kong Global Banking posting and the Graduate Talent Program page. The employer’s own listings contain the AI requirement. Vacancies and deadlines can change.
The practice exercise and worker questions are The Grind Hotline’s recommendations, not a leaked assessment or an official UBS preparation guide. The sources reviewed do not establish a reduction in junior headcount caused by this requirement. No anonymous employee discussion is presented as proof of internal policy.
About The Grind Hotline
The Grind Hotline is a two time award winning, worker first global media and workplace intelligence platform and business podcast. Its reporting and tools help people in more than 100 countries understand layoffs, AI pressure and the decisions changing their working lives.
The Host brings nearly two decades of corporate and commercial experience as an ex banker and former Fortune 100 and Fortune 500 global sales leader. He has generated revenue, trained teams and worked with senior leadership. That banking and sales background helps connect a recruitment requirement to the work behind it: analysis, client trust, targets and responsibility when the answer is wrong.
After losing his job twice in five years, he built The Grind Hotline and its three free worker tools to help employees see pressure earlier and build options. Read his account of the first firing for the experience behind that mission.
Its recognition includes the 2026 dotCOMM Platinum Award for Content Strategy and 2026 MUSE Creative Awards Silver for Branded Content, Cause/Awareness. Both can be checked through the official dotCOMM record and official MUSE record.
He is also an author, entrepreneur and founder of CallTeam, which runs outbound calling, follows up with prospects and books qualified sales conversations for B2B companies. His business work includes the 90 Day Revenue Engine for rebuilding outbound sales systems and the Sales Execution Lab for coaching teams on calls, discovery and follow up. Our Media and Editorial Standards explain sourcing and corrections. Layoff Career Counselling offers individual career support.
Important Disclaimer
This article offers reporting, analysis and general career information. Recruitment rules vary by role and location. Follow the current vacancy and assessment instructions. Neither this article nor the linked tools guarantee an interview, a job offer or continued employment.