The Missing Middle Rung: Flattening Books the Saving Now and the Shortage Later

Every delayering decision is recorded twice. Once as a saving, in the year you make it. Once as a shortage, five to eight years later, when a director seat opens and there is nobody two rungs down who has ever run a team. The two entries almost never appear on the same P&L — which is precisely why the decision keeps getting made.

Start with the span. When Gallup first measured average span of control in 2013, it was 8.2 direct reports per manager. Live Data Technologies, tracked through reporting on AI-driven delayering, puts it at 10.9 in 2024 and 12.1 in 2025. The same reporting shows manager headcount down 6.1% between May 2022 and May 2025, with executive-level roles down 4.6%.

Now do the arithmetic nobody does. Hold individual-contributor headcount constant. At a span of 8.2, a hundred ICs require about twelve manager seats. At 12.1, they require about eight. That is 8.2 ÷ 12.1 = 0.68 — roughly 32% fewer first-line manager positions to run the same organization. (The 8.2 and the 12.1 come from two different measurement programs a decade apart, so treat the figure as directional rather than precise. The direction is not in dispute.)

Read as a cost line, that is a win. Read as a capability line, it is something else entirely. The first-line manager seat is not primarily a coordination mechanism. It is the only place in most organizations where a person finds out whether they can hold a number, deliver a hard message, and be accountable for output they did not personally produce. Remove a third of those seats and you have not just removed overhead. You have cut the training capacity of the company by roughly a third, and the invoice arrives on a delay.

What the flattening actually removes

Three distinct things disappear when a management layer is compressed, and they are worth separating because only the first one shows up in the business case.

1. Coordination cost. Real, immediately visible, and often genuinely worth removing. This is the entire content of most flattening proposals.
2. Observation surface. The manager layer is where the organization watches people lead before betting a bigger seat on them. Fewer managers means fewer people observed leading, which means the succession decision four years out gets made on less evidence — the same information problem we ran the numbers on in The 11% Résumé.
3. The rung itself. A director is drawn from managers. A VP is drawn from directors. Removing the manager rung does not compress the ladder; it breaks it, because nobody steps from individual contributor to director without an intervening reps-building seat. This is the cost that surfaces last and hurts longest.

Mobility is a flow. Bench strength is a stock.

Here is where the reassuring number does real damage. LinkedIn's March 2026 Talent Trends reporting has internal mobility up 11% year over year — nearly double the 6% growth logged in 2023. Read quickly, that says the pipeline is healthy and people are moving up.

It says nothing of the kind. Internal mobility is a flow rate: how often people move. Bench strength is a stock: how many people are actually ready for the next seat. Circulating a shrinking pool faster does not deepen it. If the ladder is losing rungs, a rising mobility rate is exactly what you would expect to see — more lateral moves, more people cycling through fewer positions — right up until the moment you need a successor and discover the pool was never replenished.

The stock numbers say so directly. DDI's Global Leadership Forecast finds that 75% of organizations prioritize internal promotion over external hiring — but internal candidates can fill only 49% of critical positions immediately, and just 20% of HR leaders say they have leaders ready for their most critical roles. Eighty percent of organizations lack confidence in their bench, and DDI notes readiness has been declining since at least 2011.

A rising internal mobility rate on a shrinking ladder is not a pipeline. It is circulation.

The economy already ran this experiment

The strongest evidence that this pattern is real does not come from AI at all. It comes from a seventeen-year natural experiment with industrial robots.

Wharton's Pinar Yildirim and co-authors, working from more than 18 million résumés covering 2000 to 2017 (NBER working paper 32655), found that one additional robot per 1,000 workers reduces expected lifetime earnings by about 1.5% — and roughly a third of that decline comes not from lower pay within jobs, but from workers becoming less likely to move into better-paid ones.

The aggregate number is the one that should stop you. Across 2000–2016, rising wages added about $16,100 to expected lifetime earnings. Weaker career progression wiped out almost $12,500 of that — roughly 78% of the visible gain, erased by an invisible loss. The raise showed up in the data. The ladder that quietly stopped working did not.

Two details from that research transfer almost exactly to the org chart. First, the workers hit hardest were mid-career, six to twenty years of experience — the precise population that feeds a director bench. Second, Yildirim's own description of the mechanism: "declines in career progression often look like declines in mobility to a missing middle rung of the career ladder," pointing to fewer moves from junior roles into supervisory and management positions. A degree did not protect anyone, and the effect was strongest in manufacturing-heavy regions — which is the same structural story we traced in the manufacturing skills gap automation made worse, one level up the org chart.

The economy ran this experiment at national scale, slowly, with machines that only touched physical work. Companies are now running it inside their own org charts, faster, with software that touches coordination work. Expecting a different result requires an argument nobody has made yet.

Why it keeps happening

Not stupidity. Accounting.

The saving from removing a management layer books this fiscal year, attributable, in a named cost center. The shortage books in 2031 as a failed search, a longer vacancy, a director hired from outside at a premium. Different fiscal year, frequently a different CFO, sometimes a different CEO. No ledger connects them. An org chart is a balance sheet with a delayed settlement date, and almost nobody is holding the position long enough to be handed the bill.

Three ways to pay for a missing rung

Once the rung is gone, there is no free option — only a choice about which cost you prefer.

1. Rebuild it. Deliberately preserve or redesign first-line leadership seats as development capacity, even where AI made them unnecessary for throughput. Cheapest in total, slowest to pay off, and it requires a leadership team willing to carry a cost whose return lands after their own tenure. DDI's data says organizations with strong benches are 2.9× more likely to fill leadership roles internally and 2.8× more likely to outperform financially — but that is a five-year argument in a one-year budget cycle.
2. Buy it. Go external for the director and VP seats you can no longer grow. This works, and sometimes it is the only option — but price it honestly. External hires carry an 18–20% pay premium over internal promotes into the same jobs, underperform for roughly two years, and 46% of new hires are failed hires within eighteen months. Every rung you don't rebuild converts into a permanent external-search line item and the vacancy cost of the seats sitting open while you run those searches.
3. Bridge it. Use fractional or interim leadership to hold the seat and, more usefully, to build the people underneath it while you rebuild the rung. This is the option most mid-market companies underuse, because they treat fractional as a cost-saving measure rather than what it actually is in this context — a way to buy senior judgment and developmental supervision at the same time.

If you take the second option — and most companies will, for at least some seats — the quality of that search stops being a procurement question and becomes a structural one. You are now buying, from the outside, the leadership capability you used to manufacture internally. That purchase has to be right, because the fallback you used to have no longer exists. Which means the search has to close the information gap that made external hiring expensive in the first place: who this person reports into, what this culture actually rewards, whether the company needs transformation or multiplication, and what the seat demands at month 24 rather than month one.

A firm running a hundred requisitions cannot answer those questions. It was never built to. Ten searches understood deeply can — and when the rung underneath the seat is gone, deeply understood is no longer a preference. It is the only remaining margin for error.