In August, EY announced that its eight-week internship becomes a paid residency of eight to twelve months, with coaching and structured skills development, and that residents who convert will join as analysts rather than at the usual entry grade. The stated purpose is to build critical thinking, professional judgment, curiosity and technology fluency, on the argument that in an AI-powered world the distinctly human capabilities are the competitive advantage. The firm says it had been considering changes before generative AI arrived, and that the technology was the catalyst.
Professional services is where this surfaces first, and the reason is structural rather than a matter of these firms being unusually far-sighted. Their product is cognitive work of the kind AI handles well, they sell it by the hour so any change in how long something takes lands directly on the economics, and they track grade mix and utilisation closely enough to see the shift within a year or two rather than a decade. They are also selling AI transformation to their clients, which tends to concentrate attention on their own arrangements. So what emerges there is a leading indicator rather than a special case, and the same pressure is working its way through every organisation that employs people to do cognitive work.
The obvious reading is that AI ate the junior work and firms are now scrambling to replace what it took. That reading is half right, and the half it misses is the more useful one.
What the junior work was actually for
Junior work in professional services existed for an economic reason before it existed for a developmental one. The model runs on leverage: a small number of expensive people at the top, a large number of cheap people underneath, and a margin that comes from the difference. That work had to be done by someone, and the someone had to be affordable. The developmental account of it, that this was how you learned the trade, was laid over the top and was true enough often enough that nobody had to examine it. Some of it was genuine. Reviewing hundreds of documents does build a kind of pattern recognition that no seminar reproduces, and there is a scepticism that only forms after you have seen a dozen things that looked fine and were not.
But the volume was never set by what learning required. It was set by what the work required and what the pyramid could bill, and because those hours were already being paid for, nobody ever asked how many of them the learning actually needed, or what else those years might have been used for. Development was a by-product. It arrived incidentally, unevenly, at a pace determined by whatever came through the door, and it was called an apprenticeship because it looked like one from a distance.
What AI removed was the economics. The mundane preparatory work is the part it does cleanly and cheaply, which is why it went first, and with it went the pretext that had been carrying the development. This is not to say nothing was being taught. There has always been training, often a great deal of it, and every professional has a record of completed modules to prove it. But a catalogue of self-directed courses, however well made, completed to a deadline because completion is what gets measured, is a different instrument from being taught by someone who knows the work. Self-directed learning matters and no serious professional gets far without it, and it cannot do what mentorship does, because it cannot watch you make a judgement and tell you where it went wrong. What became visible when the grunt work went was that the mentorship half had been running on the by-product, and the part that could be counted had been standing in for the part that could not. That is why the question is not really how to replace what AI took. It is how to build deliberately what the work used to produce incidentally.
Most organisations are not built this way at all, and their version of this is worse. A manufacturer, an insurer, a bank’s operations division, a retailer: there is no pyramid, no up-or-out, no annual intake pushed through a machine on a schedule. People join, stay, and gradually absorb how the place works from those around them, and the vehicle for that absorption is the ordinary run of the work, routine parts included. Nobody calls it development and nobody funds it, because it appears to cost nothing and to happen by itself. So the same removal is under way with far less to catch it. When AI takes the routine work out of a professional firm, someone notices, because developing junior people is visibly part of how that business makes its money and there is a function whose job is to worry about it. When AI takes the routine work out of a finance team or a claims operation, there is no pipeline anyone was watching, so nothing registers as broken. The cost turns up in five years as a shortage of people who can be trusted with a judgement, and nobody traces it back.
This should have been in place regardless, and that is true across the professions rather than at any one firm. An organisation that says it develops people ought to be able to describe how, in terms of what a person will be able to do and by when, and few can, because the mechanism was never designed. The stated value was real; the system behind it was assumed. AI has not created that gap. It has removed the cover, and any firm now building something deliberate is ahead of the field rather than behind it.
Why freed time does not become development
Which is why the shape of EY’s answer is the interesting part. When I wrote about this a few months ago, my conclusion was that firms should reinvest the hours AI frees into mentoring, client exposure and deliberate practice. I no longer think that survives contact with a real organisation. Freed time does not stay free. It is reabsorbed within a quarter by delivery pressure, utilisation targets and the work that is already late, because nobody defends unbilled development hours in a month that is behind. Reallocating time inside an existing job is a plan that depends on restraint precisely where an organisation has none. What EY has done instead is change the shape of the path: a longer, paid, structured on-ramp that sits outside the delivery machine, with a higher entry grade at the end of it. Whatever else it is, it is a structural answer rather than an aspirational one, and structural answers are the only kind that hold.
Teaching is its own discipline
The harder problem is the one that follows, and it is where most attempts at this will come apart. Deciding to develop people deliberately commits an organisation to instruction, and instruction is a discipline in its own right. I spent years as an instructor in the military, where teaching adults is treated as a craft with its own training, its own standards and its own qualification, and the first thing you learn is that knowing a subject and being able to teach it are separate capabilities that happen to be found in the same person only by accident. The part that translates least well is the spine underneath it. The military maintains a body of trained instructors as a deliberate institution, and becoming one means being selected for it and then taught how to instruct, on a course with an assessment at the end. Mine ran a fortnight and carried a civilian qualification. Commercial organisations rarely go that far. The people asked to develop others have usually not been taught how to do it, and are not expected to have been.
Firms tend to assume otherwise. The expectation is that expertise transmits by proximity: put a junior next to a senior and the knowledge will travel. Sometimes it does. But the genuine experts are the most utilised people in the building, teaching is unbilled, and human patience for explaining thins out exactly when the pressure rises. So the organisation asks its scarcest people to perform a job they were never trained to do, at the moment they have least capacity to do it, and the gap gets filled by the thing that scales without anyone’s time, which is another course to complete. Underneath that sits a further layer, which is that a competency has to be defined before it can be taught. Professional judgement is not a topic. It has to be broken into what a person should be able to do, in what situations, to what standard, and then sequenced, practised and assessed. That is syllabus design, and it is a specialist discipline in its own right. Where an organisation holds that capability at all, it sits in a learning function, several steps removed from the people who hold the expertise it would need to encode.
This is not a gap that a learning and development function fills. Most organisations have one, and it does necessary work: commissioning, curating, administering, tracking who has completed what. What it is not resourced to do, and rarely attempts, is turn the people who hold the expertise into people who can transmit it. That capability sits with practitioners or it sits nowhere, which makes it a question of organisational design and how the work is arranged rather than a matter for HR to own.
An organisation that decides to take this seriously is committing to more than the announcement: leaders trained to run apprenticeships rather than merely to have them, an operating model that makes room for teaching in a way that survives a busy quarter, and someone accountable for the syllabus as an artefact that improves. Do that and the residency is the beginning of a capability. Skip it and the residency is an internship with a longer name.
The question worth putting to your own organisation has little to do with whether AI has changed what junior people should be doing. Ask instead what you would say if someone asked you to describe how the place develops judgement. Not the courses available. The mechanism: what a person will be able to do that they could not before, by when, and who is accountable for getting them there. Most of us have been carrying a belief about that and calling it a system. The useful thing about AI removing the routine work is that it has made the difference impossible to ignore.
References
EY (2026). From intern to leader: EY US introduces Career Residency program to transform entry-level professional experience. Press release, 17 August 2026. Reports the extension of the eight-week internship into a paid eight-to-twelve-month residency, entry at analyst level, and the skills the programme targets. https://www.ey.com/en_us/newsroom/2026/08/from-intern-to-leader-ey-us-introduces-career-residency-program
Accounting Today (2026). EY sets up Career Residency Program. Reports programme structure and the elevated joining grade. https://www.accountingtoday.com/news/ey-sets-up-career-residency-program