Most 2027 marketing budgets will look careful on paper. Every channel gets a line, every tool gets a line, and headcount gets a line that will probably rise, because labour is taking a larger share of marketing spend. Yet one line will still be missing in many plans: the junior seat that was never filled.
That absence is easy to miss. It does not show up as a resignation. It shows up as work that quietly stops being assigned to a beginner and starts being absorbed by software, by senior staff, or by nobody at all. Over a year, the team looks stable. Over five years, the pipeline that was supposed to produce your next managers has thinned out.
What the payroll data actually shows
The argument starts with a Stanford Digital Economy Lab working paper, revised in August 2026, built from anonymised ADP payroll records covering roughly 3.5 to 5 million employees per month through June 2026. It studies AI-exposed occupations across the wider economy, not marketing alone.
Employment among workers aged 22 to 25 in AI-exposed occupations sat about 19 percent below the level it would have reached if those young workers had kept pace with their less exposed peers. In raw terms, the two most exposed groups of young workers fell by about 11 percent since November 2022, while the three least exposed groups grew by about 10 percent. Older workers show no similar pattern. Every older age group grew over the same period, including growth of up to 11 percent for workers aged 35 to 40.
The mechanism matters more than the headline. The gap is driven mainly by reduced hiring, not by young workers being pushed out. Separations among young workers fell at least as much in the most exposed occupations as in the least exposed ones. Companies are not firing juniors at scale. They are simply not hiring them at the old rate, and the seats that would have gone to them stay empty.
Two cautions belong beside that finding. The authors describe their results as early, descriptive indicators, not proof that AI caused the shift. The size of the gap also shrinks when the analysis controls for education, and the smaller estimate is only weakly significant. Treat the 19 percent figure as a warning light, not a verdict.
Why an intact team can still be shrinking
A manager can present a team chart with no departures and no vacancies and still be running a shrinking team. If the junior hire planned for this year was never approved, the chart looks complete. It is complete only in the sense that nobody left. It says nothing about who was never invited in.
That changes how you should judge claims about AI paying for itself. A team can report hours saved, faster output, and lower cost per task while its future capacity narrows. Both facts can be true at the same time, and a budget that reports only the first one tells half the story.
It also changes promotion readiness. If entry level hiring stalls for two or three budget cycles, the pool of candidates for the next manager roles gets smaller, even if nobody leaves. You will feel that shortage later, when the obvious internal candidate was never hired.
Where junior work is moving
Not all junior work is disappearing. The research draws a useful line between tasks where AI substitutes for a person and tasks where AI complements a person. Where substitution is strong, employment falls. Where complementarity is strong, employment holds steady or grows, especially for experienced workers. The adjustment shows up in headcount, not in pay, so you cannot spot it by watching salaries alone.
For marketing teams, that line is best drawn task by task. Recurring reports, bulk title tag production, and routine bid adjustments are work a model can now handle quickly. Those tasks used to be how juniors learned: by doing the repetitive work, seeing the patterns, and asking why a result moved. When the repetitive work disappears, the learning path attached to it disappears too, unless you deliberately build a new one.
The replacement path looks different. Community work, digital PR that earns mentions and citations, and tracking where AI answer engines cite your brand depend on relationships, judgement, and context. A junior role built around that work has outputs a leader can inspect weekly: conversations started, mentions earned, citation gaps found, and follow ups completed.
A higher labour share does not automatically mean a healthier pipeline. Industry data cited in the original column shows labour rising from 21.9 percent to 24.5 percent of marketing budgets. If that extra money goes mainly to experienced hires while entry level seats stay empty, the budget has funded the present at the expense of the future.
Build a junior role a CFO can defend
The strongest way to protect a junior seat is to stop defending it as headcount and start defining it as a learning role with a budget line of its own. Put it in a capped learning bucket, with a named owner, fixed weekly hours, and a decision date.
Start by listing what your youngest team members actually do each week. Sort every task into two groups: work a model can substitute for, and work that depends on judgement, relationships, or context. Run that check yourself, even if it is only a month of simple time tracking, because the payroll research does not break out marketing roles. Your own task list is the evidence your budget needs.
Next, rewrite the junior role around the second group. Give it a weekly output you can name without vague language: a set number of hours on community participation, a list of off site mentions tracked, a citation gap report for your brand in AI answers, and a follow up log for digital PR outreach. If the role cannot describe its weekly output in those terms, it is a seat kept out of habit, and a CFO is right to question it.
Then add the structure that turns intention into a budget line: a named owner, fixed weekly hours that urgent tasks cannot crowd out, and a quarterly review with a decision date. Community work is often suggested at roughly 5 percent of a marketing budget, sometimes 10 percent, but the discipline matters more than the percentage.
A scorecard that counts empty seats
Most AI scorecards track hours saved, hours returned to higher value work, outcomes, and cost. Those measures are useful, but they are blind to the hiring gap, because an unfilled seat produces no hours saved and no cost. Add a fifth line: seats planned, seats filled, and seats deliberately left empty, with a reason for each.
Report that line beside three layers: exposure (AI citations, mentions, share of voice), behaviour (branded search, direct visits), and business outcomes (pipeline, revenue). Read them together. If exposure rises while branded search stays flat, you may have a distribution problem. If pipeline rises while junior hiring stays at zero, you may have a short term win built on a narrowing pipeline.
Make the unfilled seat line specific: the role, the date it was planned, and the reason it was not filled. A freeze can be revisited, a trade off can be rebalanced, and a claim that a model covers the work should be tested against the weekly output the role would have produced.
What to do before you sign the budget
Before you approve your 2027 plan, run a short audit. Count junior seats planned over the past two years against seats actually filled. List the tasks those juniors would have learned on, and mark which of those tasks a model now handles. Rewrite one junior role around community, digital PR, and citation tracking, with a named owner, fixed hours, and a decision date. Add unfilled seats to your scorecard, and review that line every quarter.
Then make the trade off explicit. If AI savings fund senior hires, say so. If they fund a junior seat, name the seat. A budget that names the junior seat, its owner, its weekly output, and its decision date gives finance something to approve rather than something to question. If some AI savings show up as a junior seat with a real job attached, your team is not just intact. It is still training, and that is a better measure of health than a chart with no gaps in it.

FAQ
Does the Stanford research prove AI is causing junior hiring to fall?
No. The authors describe their findings as early, descriptive indicators, and they do not estimate a causal effect of AI. The pattern is driven by reduced hiring rather than layoffs. Use it as a prompt to audit your own hiring, not as proof of a single cause.
Why does an unfilled junior seat matter if senior staff are more productive?
Productivity today and pipeline tomorrow are different measures. Senior staff can deliver more output with AI tools while the entry level pipeline narrows at the same time. The shortage only becomes visible later, when you need internal candidates for senior roles and find that the people who would have grown into them were never hired.
What kind of junior role still makes sense when AI handles routine tasks?
A role built around judgement and relationships: community participation, digital PR outreach, tracking mentions and AI citations, and following up on opportunities a model surfaces but cannot close. Give it fixed weekly hours, a named owner, and a weekly output you can inspect, such as a mention log or a citation gap report.
How much of the budget should go to community and learning work?
Suggested figures in the planning discussion run around 5 percent of the marketing budget for community, sometimes 10 percent. The exact share matters less than the structure: a capped learning bucket, one owner, protected hours, and a decision date for review. Without those, community work is usually the first line cut.
How should I report junior hiring to finance?
Add unfilled seats to your AI scorecard, next to hours saved, hours returned, outcomes, and cost. For each planned seat, record whether it was filled, left empty, and why, and review that line quarterly.
Should every team hire a junior in 2027?
Not necessarily. The right move is to make the decision explicit rather than accidental. If you choose not to fill a seat, record the reason and test whether a model genuinely covers the learning and output that seat would have provided. An explicit no, reviewed quarterly, is far safer than a seat that simply never appears in the plan.