Your org chart assumed a conversion rate that no longer holds

Look at your headcount plan for next year. Somewhere in it is a ratio: roughly one senior per three or four less experienced engineers, teams of six to eight, a manager per two teams. Nobody derived those numbers this year. They came from the last plan, which came from the one before, and originally from an organisation that had solved a different problem.

That ratio is a model, and like every model it has inputs. Two of them are load-bearing and neither is written down: what fraction of the people you hire at entry level become senior engineers here, and how long that takes.

My position: both inputs moved, the shape of the chart did not, and the resulting plan is arithmetically unstaffable. This is a delivery problem before it is a talent problem, because the band that fails first is the one that absorbs work.

The hidden model in a staffing ratio

A pyramid is a steady-state structure. It only holds if people flow upwards at roughly the rate the shape requires, and it fails quietly, because a shape that cannot be sustained looks exactly like a shape that can until the flow stops.

Three assumptions sit underneath it.

That entry-level roles produce mid-level engineers within a couple of years. That mid-level engineers produce seniors within a few more. And that the work needed to make either transition happens as a by-product of ordinary delivery, at no separately funded cost.

The third assumption is the one that carried everything. Nobody ever budgeted for training a junior into a mid, because the junior was doing useful work while it happened. The development was free because it rode on output that had to be produced anyway.

Entry roles convertA predictable fraction of entry-level hires reach mid-level here, within about two years.

Mid roles convertEnough of those reach senior within a few more, replacing attrition at the top.

Development is freeThe learning happens as a by-product of work that had to be done anyway. Never a line item.

Seniors have spare judgementEnough attention left over, after their own delivery, to supervise the people below them.

What actually changed

Entry-level developer postings are down roughly 40 percent. That number is usually read as a statement about hiring appetite. Read it instead as a statement about the input to the pyramid, and the arithmetic gets uncomfortable, because a pyramid starved at the base does not fail at the base. It fails one level up, later, when the band that should have been filled from below has to be filled from the market instead.

The second change is subtler and matters more. The work that produced the first conversion was the work that automated most completely: bounded, low-blast-radius, cheap to review. So even the entry-level roles you did keep no longer come with the development that used to be attached to them for free. You kept the seat and lost the mechanism.

And the third change is on the supervision side. Assumption four assumed seniors had spare judgement to lend. Under agentic volume they are the constraint: reviewing, adjudicating, absorbing changes. The capacity that used to fund development is the same capacity that now funds throughput, and it is being spent on throughput.

So the plan has a lower input rate, a broken conversion mechanism and no supervision budget, and the shape it produces is the same one it produced in 2019.

The chart you are staffing to

• Broad entry band feeding a broad middle
• Conversion assumed, never measured
• Development funded by ordinary work
• Mid-level hires available on the market when needed

The chart the inputs actually support

• Narrow entry band, each seat deliberately supported
• Conversion measured and named as a target
• Development funded explicitly, out of review capacity
• Mid-level hiring competitive and getting worse

The honest redesign

If you accept the inputs have moved, there are only three coherent responses. Choose one on purpose.

Buy from the market. Do not train. Staff the middle and top by hiring, accept the cost will rise, and be explicit that you are relying on other organisations’ pipelines. This is a legitimate strategy for an organisation with a short horizon or a strong employer brand. It is only indefensible when it happens by accident, which is how it usually happens.

Rebuild the rungs deliberately. Fewer entry roles than the old ratio implies, each with a named senior, a bounded area of ownership and real review on everything. Not more juniors with less support, which is the version that fails and then gets used as evidence that hiring juniors does not work. This costs review capacity, which is your scarcest resource, and pretending otherwise is how these programmes get approved and then quietly starved.

Change the shape. Accept a flatter structure with more seniors and fewer levels, and stop pretending there is a ladder. The honest version of this tells candidates the truth: this is not a place that develops people, it is a place that employs developed ones. Some very good organisations are exactly that and are better for saying so.

What does not work is keeping the old ratio, hiring the reduced number of entry-level people, giving them the work nobody else wants, and calling the result a pipeline.

The counter-signal worth watching: some large enterprises increased junior hiring in 2026 while startups cut it, which inverts the usual pattern. The plausible explanation is time horizon. If you need a hundred senior engineers in ten years, training now is rational in a way it is not for an organisation optimising for the next eighteen months.

What to measure

One number tells you whether your chart is a plan or a hope: how many people your organisation has moved from entry level to senior in the last three years. Not hired at senior. Moved.

If that number is zero, your ratio is decorative. That is fine as a decision and expensive as an accident, and the difference between those two is whether anyone has said it out loud.

Two supporting numbers. What fraction of your seniors joined at senior, which tells you how dependent you are on the external market. And how many hours of senior review capacity per week are allocated to development rather than delivery, which is usually zero and which is the actual budget line for everything above.

Where this breaks down

The conversion rate may recover in a form I cannot see. Someone who has worked alongside agents from their first week may develop judgement by a route that looks nothing like mine, and faster. I learned by writing a great deal of unremarkable code, and I am reasoning from that path, which is a genuine limit on this argument. Anyone who tells you confidently that the new cohort will convert slowly is guessing, and so am I.

The 40 percent figure is a market snapshot, not a structural finding. Postings fall in downturns for ordinary macroeconomic reasons. Attributing all of the decline to agents is overreaching, and if a meaningful share is cyclical then the input to the pyramid recovers on its own.

“Fewer, better-supported entry roles” costs the resource I have said is scarcest. The strongest objection to my own recommendation. Development consumes review capacity, review capacity is the constraint on delivery, and so the fix competes directly with shipping. I am not going to pretend that resolves cleanly. It is a real trade and it should be made explicitly, with a number attached, by someone senior enough to defend it when delivery slips.

Buying from the market may just be correct for you. I have written the buy strategy as the default failure, and for a company with three years of runway and a strong brand it is straightforwardly the right answer. The critique is of doing it unknowingly, not of doing it.

And org charts are less predictive than this argument assumes. Plenty of organisations run with ratios that make no sense and function well, because the informal structure does the real work. A model that says your chart is unstaffable is a claim about arithmetic, not about outcomes, and teams route around arithmetic all the time.

The takeaway

Your staffing ratio encodes a conversion rate and a time constant that nobody has re-derived since the inputs moved. The base narrowed, the mechanism that converted entry to mid stopped running for free, and the supervision capacity that funded it is now spent on absorbing change.

There are three honest responses and they are all defensible. The indefensible option is the one most organisations are currently running, which is the old shape with none of the conditions that made it work.

If you take one thing into next week: find the number of people your organisation has taken from entry level to senior in the last three years. Then look at next year’s headcount plan and see whether it assumes that number is larger.