In a classical kitchen the youngest cook makes the stock. It is the least glamorous job in the building: bones roasted and covered with water, onions and carrots cut to a size no guest will ever see, a pot watched for eight hours so that it never quite boils, then skimmed, strained and reduced. Nobody orders stock, and nobody tastes it as itself. The commis who makes it is paid little and taught a great deal, because stock is where a cook learns what a sauce should taste like before anyone lets them near one, and a cook who has made a thousand gallons of it can tell from a single spoonful of a finished plate that the base underneath was rushed. The kitchen does not keep the commis for the stock. It keeps the commis because in ten years that person is the sous chef.
Every kitchen eventually discovers that stock can be bought. The bases come in tubs now, they are good, and a restaurant that switches saves a salary and eight hours of stove a day. The cost of the switch never appears that year. It appears a decade later, when the chef goes looking for a sous chef and finds that nobody in the building has ever made the thing everything else is built on.
Marketing has started buying the stock.
The Bottom Rung Is Thinning
The numbers arrived this week from several directions at once. Northeastern University's AI and Society survey, published Monday, asked 8,174 hiring decision makers how AI had changed their plans, and about half said it had. Among those, 26% are hiring fewer entry level candidates while 7% are hiring more, and 28% now expect more output or more senior work from the juniors they do bring in. Software already screens résumés for more than half of the respondents. The Content Marketing Institute found the same shape inside marketing departments this spring, with 34% of marketers reporting less entry level hiring against 14% reporting more, and Pave compensation data cited in that report shows entry level roles falling from 8% to 4% of teams in two years. Economists at the Census Bureau, studying the industries most exposed to AI, found employment among 22 to 24 year olds down 12% over ten quarters, driven by fewer hires rather than by layoffs.
None of this is irrational at the level of a single team. The junior marketer's traditional work was the stock of the department: first drafts, competitor scans, campaign builds, reporting decks, the hundred small tasks through which a person learns what good looks like. Software now produces a passable version of most of that work in seconds. A manager looking at a headcount request is asking a fair question when she asks what the new hire would do that the tools do not.
The Subsidy Nobody Priced
The honest answer is that the junior seat was never mainly paying for the output. It was a bargain in which the company got cheap work and paid part of the wage in teaching, and the bargain held because the work was real enough to cover the cost of the lessons. Gary Becker drew the distinction that matters here more than sixty years ago: training that is only useful inside one firm is easy to pay for, while training in general skill is hard to fund, because the worker can carry it across the street. Taste in marketing is about as general as a skill gets. The old arrangement survived only because the drafts and the decks quietly paid for the education.
Take the drafts away and the education shows up on the books as pure cost, and pure cost spent on general skill is exactly what markets underfund. Every firm would rather hire the person somebody else trained. CMI found that 28% of marketing teams have already hired overqualified candidates into roles below their experience, which is the restaurant version of hiring away the sous chef from the place down the street. It works for the first kitchen that tries it. It stops working when every kitchen tries it, because by then nobody is left making stock.
The Agents Need Tasters
The irony is that the same tools are raising the price of the thing the junior seat used to produce. Last month Anthropic published a study of how employment has changed inside its own company, a business further along this curve than any marketing department. The junior share of its technical staff fell from 21% to 14%, and juniors fell from 27% of technical hires to 9%. By August the agents its people direct were logging about six hours of work for every hour of human work, and simply reading what those agents produced would take roughly half of an employee's week. The authors conclude that human attention and judgment have become the constraint, that hiring managers now screen for exactly those qualities, and that staff are drifting away from the hands on work that built their skill in the first place.
That is a kitchen with a hundred tubs of bought stock and three people who can taste. Output has become cheap and judgment has become the bottleneck, and judgment is the one input the industry has stopped manufacturing. Virginia Ford-Young, chief marketing officer of S&P Global Ratings, made the requirement plain in the CMI report: you need real marketing experience to know whether a piece of work is any good. The question for the next decade is where that experience is supposed to come from if nobody starts at the bottom.
Teach the Palate
The useful response is not to protect the old job. Nobody needs a junior marketer to format a reporting deck by hand again, any more than a modern kitchen needs a commis to churn its own butter. The response is to keep the seat and change what it is for. A junior hire in a department full of agents should be learning to taste: reading what the agents produced against the brief, checking every claim against its source, comparing three versions of a landing page and saying which one a buyer would trust, then sitting beside a senior person who explains why the call was wrong when it was. That is training in judgment, and it is precisely the work Anthropic's own numbers say a department runs short of. Some of it should still involve making the thing from scratch, often enough that the palate has something to calibrate against, because a cook who has never made stock cannot tell when the bought base has turned.
The economics argue for doing this on purpose, with its own line in the budget, rather than leaving it to the margins of a busy quarter. A chief marketing officer who treats the junior seat as an investment in future senior capacity can defend it in the planning meeting the way she would defend any asset with a long payback. One who treats it as a cost to be optimized will win this year's comparison and lose the one in 2032, when the people able to supervise ten agents are the scarcest hires in the market and the competitor who kept training them has a bench. Gary Friedman of Seyfarth Shaw, quoted in the Northeastern release, described the choice as a bottom rung that is either hollowed out or turned into an escalator. Which one a department ends up with is a budgeting decision, and it is being made this winter.
The bases in the tubs will keep getting better, and kitchens should use them. But somebody in the building has to know what the stock is supposed to taste like, and the only way anyone has ever learned that is by making it.
