AI and the office of the future are usually discussed in terms of which jobs might disappear. Far less attention is given to a more complicated question. If AI removes much of the junior work through which employees traditionally gained experience, who will train the managers and senior professionals of tomorrow?
McKinsey examines this problem in its July 2026 report, Building Expertise in the Age of AI: Who Trains the Next Generation? It warns that research, documentation, data preparation, basic coding and preliminary analysis are increasingly being absorbed into AI systems.
These tasks may appear routine, but they served another purpose. They allowed inexperienced employees to observe senior colleagues, understand how organisations operated, make relatively safe mistakes and gradually develop professional judgement.
McKinsey believes businesses must redesign training to preserve this development. But its report also leads to a more uncomfortable question. What happens if the office of the future needs far fewer managers and senior administrators anyway?
The traditional career ladder
The conventional office was built like a pyramid. Large numbers of administrators and trainees occupied the bottom. Above them sat experienced employees, followed by managers, directors and partners.
People generally progressed by completing increasingly difficult work. A junior finance employee might begin with reconciliations, basic reports, document preparation and routine client administration. Over time, that employee would encounter unusual cases, learn from mistakes and become trusted with more important decisions.
This was not a perfect system. Progression was already uneven. Surveys conducted in the past found that graduates were often perceived by business leaders to progress faster than employees without degrees.
One reason is that graduate schemes are deliberately structured to accelerate development. Participants receive formal training, departmental rotations, support with professional qualifications, mentoring and earlier access to senior management.
Meanwhile, somebody entering through an ordinary administrative position may spend years acquiring considerable practical knowledge without being offered the same opportunities. In some organisations, an academically qualified employee can progress rapidly towards senior management while a highly experienced colleague without a degree remains at manager level.
A degree does not automatically make somebody a better manager, and promotion also depends upon performance, professional qualifications, personality and opportunity. Nevertheless, graduate programmes often create a faster and more visible path to leadership.
What happens when the first steps disappear?
AI could make this unequal career ladder even narrower.
If research, administration and preliminary analysis are automated, businesses may recruit fewer people to perform them. This means fewer opportunities for somebody to enter at the bottom, demonstrate ability and progress through practical experience.
The first evidence may not be a dramatic redundancy announcement. It could be a vacancy that is never advertised, a retiring employee who is not replaced or a graduate scheme reduced from ten positions to five.
We recently examined eight finance companies operating within one offshore finance centre. These were international businesses with offices in several countries, allowing us to look beyond their local operations and examine what they had said about AI elsewhere. Each had publicly discussed significant investment in AI, either in company accounts, corporate reports, public statements or media coverage concerning other parts of its international business.
This does not necessarily mean that every AI system was already being used within the offshore centre itself. However, it demonstrates that these companies were actively investing in AI across their wider organisations and were familiar with its potential to automate professional and administrative work. At the time of our snapshot search, only one trainee position was being advertised across all eight companies.
Although vacancies change and some positions may be filled internally or advertised elsewhere, finding only one trainee opportunity across eight major finance employers remains striking. It appears far removed from the traditional offshore model of regular trainee recruitment and raises an important question: are these businesses still planning to employ large numbers of trainees, or are they already anticipating a future in which fewer people are required?
McKinsey’s proposed answer
McKinsey argues that businesses should replace informal workplace learning with a more deliberate system.
One suggestion is the answer key model. A trainee completes an assignment independently before comparing the result with work produced by AI. A manager then discusses the differences, helping the trainee understand what was missed and why one conclusion may be stronger than another.
The report also describes companies using curriculum based roles, structured assignments and simulations. Bank of America, for example, is using simulated experiences to accelerate the development of judgement that junior bankers previously gained from routine work.
McKinsey refers to the medical residency model as another possibility. A small group of junior employees could work under the direct supervision of an experienced professional until they demonstrate the ability to make decisions independently.
These ideas are plausible. AI could create realistic clients, financial problems, regulatory investigations and difficult management situations. Trainees could practise responding to them without placing real customers, money or businesses at risk.
However, this approach creates a substantial cost.
Who pays for training?
In the traditional model, trainees learned while also producing useful work. Their research, documentation and administration contributed to the organisation, even if it required checking by somebody more experienced.
If AI completes that work faster, more cheaply and more accurately, training becomes a separate expense. A company may have to pay the trainee, build simulations and take experienced managers away from productive work to provide coaching.
McKinsey acknowledges this problem. Why should one business bear the cost of training somebody who might later leave for a competitor?
Its answer is that company training would use the organisation’s own knowledge, cases and systems. This would make at least part of the employee’s expertise specific to that business. McKinsey also warns that companies refusing to invest could eventually face a shortage of experienced leaders.
That assumes, however, that businesses will still need a substantial pipeline of future leaders.
Will the future office need as many managers?
Much of the discussion about AI assumes that junior jobs will decline but large numbers of managers and senior professionals will remain. That may underestimate what AI agents could eventually do.
Management includes allocating work, monitoring progress, checking compliance, reviewing performance, preparing reports, identifying problems and passing information between different parts of an organisation. Many of these activities are already within the developing capabilities of AI.
The same applies to senior administration. AI systems could monitor thousands of transactions, examine correspondence, maintain records, identify anomalies, prepare regulatory reports and refer only exceptional cases for human attention.
The future office may therefore need fewer people at several levels, not only fewer trainees.
Instead of a traditional department containing administrators, junior professionals, supervisors and several layers of management, a business could employ a small group of professionally qualified people supervising teams of specialised AI agents.
This is sometimes described as swarm AI. Rather than relying upon one system to complete everything, multiple AI agents divide a complicated process between themselves. One agent researches, another analyses, another checks compliance and another reviews the final output. A human professional receives the completed result and becomes involved when judgement, responsibility or personal communication is required.
This remains a possible scenario rather than a certain prediction. AI reliability, regulation, legal responsibility, confidential data and public trust may all slow its adoption. However, the commercial incentive is clear. A small team of highly qualified professionals supported by AI could potentially produce the work of a much larger office.
A much more selective route into finance
If this happens, businesses may recruit only a small number of trainees because they will need only a small number of future managers.
Competition for those positions could become intense. Companies may use degrees, advanced professional qualifications and academic results as convenient filters when hundreds of people apply for a handful of opportunities.
The successful candidates could be placed on highly structured programmes and rapidly trained to supervise AI systems. They might progress much faster than previous generations because AI gives them immediate access to knowledge, examples and feedback.
Others could find that there is no longer an ordinary route into the profession. The administrative position through which somebody once gained experience may have disappeared. The opportunity to prove ability through years of practical work may no longer exist.
AI could consequently make professional careers more academically selective. A small group of highly qualified graduates might receive exceptional training and rapid progression, while those without degrees or established professional qualifications struggle to enter the industry at all.
The problem would not simply be fewer jobs. It would be reduced social mobility.
Could training move outside the office?
If companies become reluctant to pay for large trainee programmes, some professional training may move into universities, industry academies or specialist education centres.
Students could complete realistic workplace simulations before applying for jobs. They might manage fictional clients, investigate suspicious transactions, review accounts containing deliberate errors and defend decisions before an AI examiner.
Professional bodies could assess judgement and responsibility rather than relying mainly upon written examinations. Employers would then recruit people who had already demonstrated some practical ability in simulated environments.
McKinsey does not predict that dedicated training centres will replace workplace experience. It discusses internal simulations, external work with charitable organisations and residency style programmes led by employers. Moving much of professional development outside the workplace would take its argument considerably further.
It also raises another problem. Simulated experience cannot fully reproduce the consequences of making a real decision involving real clients and real money. Someone must eventually be trusted with genuine responsibility.
Who trains tomorrow’s managers?
McKinsey is right to warn that companies cannot remove junior work without considering how future expertise will be created. Its training proposals could help businesses preserve professional development while AI absorbs routine tasks.
But the office of the future may require fewer future managers in the first place.
Over the next five years, we could see smaller trainee intakes, fewer administrative positions and fewer management layers. A select group of professionally qualified people may supervise teams of AI agents capable of completing much of the work previously distributed throughout an office.
The transformation may occur quietly. Industries could remain profitable and successful while employing considerably fewer people. As AI is adopted internally to perform particular functions, this may be the point at which redundancies begin or departing employees are no longer replaced. Companies may never announce that AI has replaced hundreds of jobs. They may simply stop creating those jobs.
The question is therefore larger than who trains the trainee. It is who will still be selected for training, who will pay for it and how many human managers the AI enabled office will ultimately need. This could prove to be the biggest shake up of careers in the finance industry for at least 50 years.
Published: 14th August 2026