The junior base of the org chart is being removed a headcount at a time, and almost nobody is framing it as a decision. On a trading desk, where the shape was already close to a diamond, that lands as acceleration rather than disruption.
A note from Matt. The subject this month is the shape of the org chart, and why trading desks are going to feel the change before most people do. Pete Hirsch's piece on the collapse of the pyramid and the rise of the diamond set me off. It is about corporates in general. I want to pull the trading thread out of it.
The base is being removed
Ask a desk head how many graduates they took on this year against five years ago. The number has usually gone one way, and almost nobody has framed it as a decision. It gets made a headcount at a time.
The hard numbers on this are from the US. Indeed's Hiring Lab reported last month that as of May 2026 senior-level postings in the US were up 14.7 per cent year on year, while entry-level postings were down 7.5 per cent and have been falling since they peaked in 2022. The report was titled "The Labor Market Is Tilting Toward Seniority". Stanford's Digital Economy Lab found a 16 per cent relative decline in employment among 22 to 25 year olds in the occupations most exposed to AI.
Elsewhere the picture is different, and in places it points the other way. In Singapore the Ministry of Manpower has the share of fresh resident graduates in work by June rising, from 47.9 per cent in 2024 to 51.9 per cent in 2025. But shifts like this tend to surface in the US first and land in other markets years later, by which time the hiring decisions that caused them are long made. That is the only useful window a leading indicator ever gives you.
Hirsch's argument is that what remains is not a smaller pyramid but a different shape: a diamond, with a slim junior base, a fat middle of experienced specialists and a slim top. The fat middle is still an argument rather than a measurement. Indeed has mid-level postings down 6.7 per cent against January 2025, so in the aggregate the middle is not visibly thickening yet (I think largely down to across-the-board AI-related headcount recalibrations). The long-term direction of the base, however, is not in question.
Trading desks will, I expect, be among the first places we see the diamond solidify, because they were already amongst the closest things to it in the corporate world. A handful of desk heads. A deep core of traders, originators, analysts, risk, structuring and trade finance. A thin layer of trade support beneath. What hits other industries as disruption hits trading desks as acceleration.
The middle is where the war is: relationships will reign
Whether or not the middle thickens in the aggregate data, on a trading desk it is the band that runs the money and carries the relationships, and it is the band everyone is competing for, because nobody can grow it cheaply from below at the old rate.
The obvious objection is that if software can absorb the junior work it will eventually absorb the middle too, so why pay up for it. I do not think that follows, for a reason specific to this market. A great deal of the value in commodities sits in relationships that took years to build and cannot be documented, transferred or modelled.
The broker who returns your call at seven in the morning because you dealt with them fairly the time it went against you. The originator who sees the cargo first because a counterparty remembers how they behaved when performance got difficult. The trader whose read on a market comes from twenty years of conversations rather than twenty years of data. None of that sits in a system. It sits in a person, it leaves the building when they do, and it is most of why a desk is worth more than the sum of its screens.
The tools have taken the work that could be written down: the model build, the first-pass screen, the summary. What is left is the work that was never written down. Pricing an unusual cargo. Judging whether a counterparty will actually perform. Knowing who to call. Automate the analysis and you do not thin the middle. You concentrate its value.
What good looks like has narrowed
So the specification is shifting. People who can direct the tools rather than compete with them, which means knowing which question to ask and being able to smell a wrong answer. People who can carry more ground with less headcount support beneath them, because the support is not coming back. And, increasingly the thing that separates two good candidates, people who can build and hold the relationships above.
The on-ramp is a ten-year decision
Here is the talent arithmetic that worries the leaders I speak to. Someone genuinely useful in the middle of that diamond has usually seen eight to ten years, which in this market means at least one full cycle and one proper dislocation. Judgement here is built by being wrong in public and having to carry it. You cannot compress that and the tools do not teach it.
So a house that stops hiring juniors this year is not making a decision about this year. It is making a decision about 2035, and by then it will have forgotten it made one. The cost turns up as an empty bench a decade later, at which point the only option left is to buy experience at whatever the market is charging.
The contrarian bet
The organisations that keep a junior bridge open while everyone else closes theirs will have a middle to promote from in ten years, at a fraction of what their competitors will be paying to buy one.
In some instances, the case for cutting is legitimate. Training juniors is expensive, most of them leave, and the manager who pays for it is rarely the one who collects.
"Most of them leave" is the serious objection, and I think we've already touched on the answer. You are not trying to keep all of them, only the few who'll become the linchpins of the middle layer. And the ones who leave do not disappear. They turn up at counterparties, at clients, on the other side of your trades. In a business built on who returns your call, a decade of alumni spread across the market is not the cost of training. It is the network.
Which is why so few will do it. And an uncrowded trade is usually the one worth having.
On my radar
One to actually do something about. If the two or three people who carry your key relationships left this year, what leaves with them, and is any of it written down anywhere? Most desks I speak to have never fully mapped it. They can tell you their VaR to the decimal and could not tell you which counterparties would stop returning calls. It costs nothing to work out and it is the single most concentrated headcount-related risk on a lot of desks.
Whether desk heads treat the junior layer as a cost to trim or a pipeline to protect. The answer tells you who is thinking past the next book.
Comp pressure building in the middle tier as the buy-not-build reflex takes hold.
Next month I want to stay at the base of the diamond and ask a harder question. Not how many people we hire in, but whether they should all be the same. Graduate schemes across this industry have converged on one very narrow specification, and I am not convinced that is good risk management.
Sign-off
If you are hiring anywhere across commodities trading and the commercial, functional and leadership roles that surround the desk, that is exactly where I spend my time, and I am easy to reach.
The piece that set this off: Pete Hirsch, The Collapse of the Pyramid & the Rise of the Diamond.