The Pyramid Meets the Clock

The COO asks who owns the decision to change service policy on the lanes that are bleeding margin. The head of sales says the customer escalation path runs through a regional president who is out this week. The plant manager says changing the production sequence will violate a quality hold rule that was put in place after last year’s audit. The CIO says the data is good enough to point, not good enough to bet the quarter on without a sign off from finance. The CFO says the sign off will not happen until the monthly close locks. Everyone is acting responsibly. Nothing moves. The meeting ends with a calendar invite. The problem is not that nobody knows what to do. The problem is that permission takes longer than the business can afford. In late 2025, news that McKinsey has discussed cutting roughly ten percent of some non-client facing teams landed like a management story, then kept landing like something else. It reads, if you are willing to see it, like a timing story. Bloomberg reported the discussions as a response to a consulting slowdown, with cuts concentrated in support functions. Fast Company framed it as a warning sign for consulting in the AI age, less about one firm and more about the business model. The surface is layoffs. The deeper issue is that the consulting pyramid was built to sell analysis time, and analysis time is being priced toward zero. This piece is about what happens when the cost of producing answers collapses, but the cost of acting does not. It is also about why “Power of One” narratives are naturally attractive, and why they can be descriptively true while still leaving a CEO with nothing to change on Monday morning. McKinsey Global Institute’s “The power of one: How standout firms grow national productivity” argues that national productivity growth is driven by a small set of large firms taking bold strategic action. The question is not whether their decomposition is defensible within the sample. The question is what CEOs should do with a story that is optimized for clarity rather than control. What would have to be true for this outcome to keep repeating.

The story that sells because it feels like science

The McKinsey Global Institute report is careful in its framing, but the headline is still built for the human brain. A few firms matter. Standouts move the needle. Most firms do not. That kind of statement carries emotional force because it compresses complexity into a plot. It gives leaders a hero-class to study, and it gives boards a sense that productivity is not a fog. It is a function of decisive actors. The report also makes choices that make the narrative cleaner. It limits the analysis to roughly 8,300 large firms in Germany, the UK, and the US. It excludes micro, small, and medium sized firms and startups. That exclusion does not make the work wrong. It makes it conditional. When you remove the long tail, you concentrate measured contribution in the remaining head. That mechanical fact does not invalidate the findings. It does mean the findings are about a constructed universe, not the whole economy.

Then the metric choice does something even more important. They use an economic definition of productivity as real gross value added per worker. In the technical materials, they define gross value added at the firm as EBITDA plus personnel costs, then deflate using sector level deflators because firm level prices are not available. That is a legitimate approach in productivity literature when you are trying to compare across firms. It also means the “productivity” being measured is closer to value creation per worker than to operational throughput. Pricing power, portfolio mix, capital intensity, and accounting structure can show up in the number. This is why operators sometimes read the conclusion and feel a misfit in their bones. A plant manager reads “productivity” and thinks about conversion, yield, downtime, and variation. A CFO reads “value added per worker” and sees a measure that naturally rewards firms that move into higher margin categories, gain pricing, or reallocate resources toward advantaged activities. Those are real drivers of prosperity. They are not the same thing as doing the current work better. The report then decomposes the growth into components, using methods that credit both within firm productivity changes and reallocation of activity toward more productive firms. Reallocation is not an accounting trick. It is part of how economies get richer. It also creates a narrative that naturally favors scale. Firms that win share and hire more workers will, by construction, matter more to the measured growth. That is not a moral judgment. It is how the math works. So the story can be true and still fail as guidance. It tells you who won in a defined period. It does not tell you how to build the decision system that would allow you to win under different conditions. That gap between description and control is where boards get misled. Not by lies, but by the comfort of a clean narrative.

Consulting’s real product was permission, not answers

To understand why McKinsey layoffs matter, you have to be honest about what consulting sold at its peak. It sold analysis, yes, but analysis was never the final product. The final product was permission. Permission to act, permission to cut, permission to invest, permission to exit a business that had become a shrine. A board could do something politically costly if an external firm had said it. A CEO could move against a powerful internal constituency if a benchmark justified it. A CFO could demand standardization if an outside view declared the variance intolerable. The slides were a transport mechanism for consent. That model required time. It required time for interviews, time for workstreams, time for synthesis, time for alignment sessions. The time was not a bug. It was the engine. While time passed, the organization adjusted emotionally. People became resigned. Alternative coalitions formed. Decisions became less personally threatening. The outside firm was a neutral third party who made delay look like diligence. The consulting pyramid was built to monetize that time. Junior labor produced drafts and analysis. Senior labor provided judgment and political cover. Billing turned time into revenue.

The client accepted the trade because the cost of internal coordination was high and the cost of external validation felt lower than internal conflict. Then AI arrived and did the obvious thing. It compressed the production of analysis. Fast Company’s point, stripped of the rhetoric, is that the source of value in consulting is moving away from information synthesis because AI can do much of it. The Financial Times has described how AI threatens the traditional “pyramid” model by reducing the need for large cohorts of junior analysts, even as firms keep pay high. Bloomberg’s reporting on the McKinsey cuts emphasizes non-client facing departments, which is where the coordination machine lives. If analysis can be produced with fewer hands, then the overhead built to manage those hands becomes friction. This is not a story about intelligence being replaced. It is a story about delay losing its cover. When analysis becomes abundant, the organization’s inability to act becomes visible. The bottleneck moves from cognition to authority. Once that happens, many engagements that used to look like “strategy work” begin to look like what they always were. Paid time inside the permission staircase. The permission staircase is the motif most executives recognize and least want to name. It is the series of adjacencies an issue must cross before a decision can be made. Each step is defensible. Each step has a reason. The cost is that the enterprise pays for time as if time were free.

The hidden mechanism is latency, and it is falsifiable

“Latency” is a word people associate with networks, but the enterprise has it too. The simplest definition is the elapsed time between signal and changed behavior. If you can measure it, you can manage it. If you refuse to measure it, you will keep confusing activity for progress. The mechanism is falsifiable. Pick a business outcome that matters, something that shows up in cash, margin, safety, quality, service, or retention. Trace the signal that indicates drift. Then measure the time from the first detectable signal to the first corrective action that changed the trajectory. Not the time to create a slide, not the time to convene a meeting. The time to change behavior. Most firms never run this measurement because it is politically dangerous. It reveals where decisions actually happen. It reveals where they do not happen. It reveals the people and committees who function as gates, and it reveals that those gates often exist to manage blame rather than manage the business. AI does not solve this. It exposes it. A CFO can dismiss a thousand claims about cultural change. A CFO cannot dismiss elapsed time when it shows up as working capital that should not be there, as expedited freight that never stops, as margin leakage that repeats every quarter. Time becomes money in plain sight. When

service failures persist, you pay in retention. When quality escapes repeat, you pay in scrap, warranty, and credibility. When safety warnings do not convert into changed conditions, you pay in injuries, and sometimes you pay in lives. Consulting’s old model was comfortable because it let everyone pretend the problem was knowledge. If the problem is authority, you cannot outsource it. You can hire help, but you still own the design. That is why the economics are changing. The highest value work is not analysis. It is redesigning how permission moves, and converting that redesign into behavior. This is also why “Power of One” narratives land so easily. They offer a story in which winners make bold moves and everyone else should imitate those moves. It moves the discussion away from the hard question, which is why most firms cannot move at all. It replaces control with admiration. You can see the difference in how the two arguments treat time. The “Power of One” story treats time as a backdrop. The consulting story treats time as billable. The operator’s story treats time as the variable that decides whether a signal becomes a result.

Counterevidence that deserves respect

There is a credible counterargument. Human judgment is still scarce. Relationships still matter. Boards still need trusted intermediaries. Some decisions are expensive and irreversible, and speed can be a form of recklessness. This is true. If your firm is about to divest a core business, or acquire a competitor, or renegotiate a union contract, you do not want a machine hallucinating a strategy. You want people who have seen similar situations, who can sense second and third order effects, who understand how incentives bend behavior, and who can detect when a clean model is hiding a messy reality. The counterargument becomes wrong when it is used to protect delay. Judgement does not require months of synthesis. It requires clear ownership, clear evidence, and clear authority to decide. When organizations use the rhetoric of prudence to defend permanent handoffs, they are not buying judgment. They are buying time, and time is the one asset they cannot replenish. This is why the layoffs are not just a cyclical correction. They are a pressure release in a model that assumed time would remain billable. When AI reduces the time to produce analysis, the old pyramid loses its economic rationale. The remaining work becomes closer to operating design and implementation. That pushes consulting firms toward different staffing patterns, different pricing, and a different relationship with clients. The counterargument also matters for how you read the “Power of One” report. Descriptive work can be valuable. The problem begins when a descriptive decomposition becomes a prescription. The report’s claim that fewer than 100 firms can account for a large share of measured productivity gains is plausible within their sample and metric. It does not tell you what to change inside your decision system. It does not tell you how to build the control capacity that allows you to act without constant re litigation.

Re litigation is the enemy of conversion. Conversion is the work of turning signal into action, and action into improved outcomes, with minimal delay and minimal drama.

Two diagnostic paragraphs boards can read aloud

If your enterprise had to respond to a material service failure this week, how many explicit sign offs would be required before the corrective action that actually changes behavior can be executed, and how many of those sign offs exist because the person granting them fears being blamed if the decision goes wrong? When the organization misses a quarter and everyone agrees the cause was visible in advance, where exactly did the signal first appear, and what was the measured elapsed time between that first appearance and the first irreversible action that would have changed the result? Those are not rhetorical questions. They are audit questions. If you cannot answer them with dates, names, and timestamps, you do not have a strategy problem. You have a control problem. Control is not command and control theater. Control is the ability to intervene inside guardrails with speed, and to learn from the intervention. The reason AI matters is that it makes intervention possible at scale. The reason AI does not automatically pay is that most firms have not redesigned the permission staircase so intervention can happen without endless re litigation.

Why “Power of One” feels fitted to a sale

This is the part you can say with confidence without claiming motive. Their result is sensitive to scope and definition choices that they made explicitly. Those choices naturally produce a clean narrative that is easy to sell. Large firms only, three countries, four sectors, 2011 to 2019. Productivity defined as real gross value added per worker, constructed from EBITDA plus personnel costs, deflated at the sector level. Decomposition that credits reallocation toward more productive firms as part of national gains. A standout class defined by outsized contribution, which will always produce concentration because the class is created based on contribution. None of this is inherently wrong. It is a set of boundaries. Boundaries matter because they determine what competing explanations are allowed to survive. When you remove the long tail, you remove the part of the economy where diffusion, entry, and small firm experimentation are the story. When you choose a value creation measure, you naturally favor portfolio moves and pricing power as drivers. When you credit reallocation, you turn scale into a contributor. If you come from operations, you tend to carry a different standard. You want to know what to change in the causal mechanism, not just what correlated with winning. You want a model that tells you how to convert signal into action with less latency. The “Power of One” report, read as a story, risks sending leaders into admiration when they should be redesigning control.

This is also why consulting loves these narratives. They function as justification for advising bold strategic moves. They also carry a subtle message that only a small set of elite firms can do it, which preserves the prestige of the advisor class. In a world where analysis is cheap, prestige is not enough. The buyer will ask for conversion. The buyer will ask for control. The buyer will ask for reduced elapsed time.

What the consulting layoffs are really telling you

Bloomberg reported that McKinsey leaders discussed cutting roughly ten percent of headcount in some non client facing departments. That is where much of the coordination machine lives. Fast Company’s framing is that this is a warning sign for consulting in the AI age. The Financial Times has described how AI threatens the pyramid staffing model, with firms freezing starting salaries and hiring fewer junior analysts because the work can be done with fewer hands. Put those together and the picture becomes clear. Consulting firms are being pushed to do what they have often advised clients to do. Remove layers that exist to coordinate work that no longer requires coordination at that scale. Compress the handoffs. Reduce the distance between signal and action. Make the machine lighter. The irony is that consulting firms can do this faster than most clients because their permission staircase is shorter. They can change staffing models with a partner vote. Most clients cannot change decision rights without a cultural war. That is why the consulting industry will not die. It will tighten around what clients cannot do alone. Political cover. Operating design. Implementation. High stakes decision support. The part that dies is commodity analysis billed as labor. You can see the next form emerging already. Firms will push more work into technologyenabled delivery. They will price more on outcomes, or at least claim to. They will hire fewer generalist analysts and more specialists who can build systems. They will sell “capability building” that is really permission redesign, because clients cannot admit they have permission problems. This is where your argument has teeth. You are not arguing that consultants are wrong about standouts. You are arguing that the primary determinant is the control layer. If your permission staircase forces every intervention into re litigation, you will not become a standout by copying surface moves. You will become a slightly busier version of yourself.

A prediction worth being embarrassed by

By the end of 2027, the market for pure strategy decks sold on hours will be materially smaller than it was in 2019. Firms that remain dependent on junior leverage for synthesis will either reduce intake, compress their pyramids, or move toward productized delivery. Buyers will pay premiums only where the engagement changes decision latency in the real system, not in a binder.

If this is wrong, it will be because buyers decide to keep paying for time they do not have. That is possible. It is also how industries lose.

The closing pressure point

The consulting model is not collapsing because smart people stopped being smart. It is collapsing because time stopped being billable. The “Power of One” story is not seductive because it is false. It is seductive because it is clean. The enterprise does not need a clean story. It needs control. If your system cannot convert signal into action without re litigation, you do not have a productivity problem. You have a permission problem.

References

This piece relies on current reporting on consulting economics and staffing, including Bloomberg’s December 2025 coverage of proposed McKinsey cuts focused on non client facing departments and Fast Company’s December 2025 framing of those cuts as a warning sign for consulting’s model in the AI era. It uses McKinsey Global Institute’s May 6, 2025 report The power of one, and its accompanying PDF and technical appendix, for the defined sample of roughly 8,300 large firms, the 2011 to 2019 window, and the construction of labor productivity as real gross value added per worker derived from EBITDA plus personnel costs and sector deflators, along with the decomposition logic that credits reallocation effects. It draws on the Financial Times’ late 2025 reporting on AI pressure on consulting’s pyramid staffing model as industry context. For the deeper mechanism, it leans on Herbert Simon’s bounded rationality in the 1950s as the foundation for why procedure substitutes for judgment under load, Alfred Chandler’s Strategy and Structure in 1962 for how organization design constrains execution, James March’s work on decision processes and institutional behavior for how responsibility diffuses through governance, W. Edwards Deming’s Out of the Crisis in 1986 for treating delay as systemic cost rather than individual failure, Stewart Myers’ real options framing in the 1970s and 1980s for the link between timing and value, and Judea Pearl’s causality work for the distinction between prediction and intervention that separates descriptive winner stories from controllable operating change.