The Metric That Tells You When the Future Has Already Moved On

The spreadsheet is not the problem. It is the receipt. This is what the next version of this article has to be about. Not whether a company is buying tools, funding pilots, or sounding fluent in the language of AI. Not whether a plant can still run, a quarter can still be defended, or a dashboard can explain what went wrong. The harder question is whether the operating model now requires materially less human effort to produce the outcome, hold the outcome, and improve the outcome. If the answer is no, then what is being called transformation may be little more than better instrumentation wrapped around the same dependency. What would have to be true for this outcome to keep repeating. Quite a lot, actually. The enterprise would have to keep mistaking visibility for control. It would have to keep mistaking activity for redesign. It would have to keep paying people to compensate for weak architecture, then calling that compensation discipline. It would have to believe that because the business can explain more of what is happening, it must therefore be getting better at changing what happens. That belief has protected a great deal of spending, a great deal of hierarchy, and a great deal of self-regard. It is getting much more expensive to defend.

The mistake most companies still want to call progress

The prevailing belief is easy to understand because it was reasonable for a long time. Enterprises are large. Systems are fragmented. Risks are real. Quality matters. Safety matters. Customers can punish overreaction just as easily as they punish delay. In that world, more dashboards, more reporting, more coordination, more analytics, more approvals, and more system coverage can feel like maturity. Boards need evidence. Management teams need milestones. Operators need traceability. It is not irrational to look for proof of change in what was installed, what was used, and what was discussed. The trouble begins when those proofs become substitutes for the thing they were supposed to serve. A planning system can be called modern while planners still spend their mornings correcting its answers. A plant can be called data rich while supervisors still carry the real model in their heads. An AI program can be called successful because it produced excellent summaries while the same people still sit in the same meetings deciding the same exceptions after the summary appears. The machine may look newer. The burden is not. That is the reason the future version of this article cannot stop at the original metric. The old test was already exposing something true. If labor burden does not fall, the business has not changed. But friction is now being repriced more directly than before. Cheap reasoning changes what the market is willing to keep funding. It changes what customers will tolerate. It changes what buyers will view as durable advantage and what they will increasingly read as trapped payroll, slow closure, and middle layers that exist because the architecture still cannot hold the logic, the permissions, and the evidence required to act.

The next era will be harder on firms that can explain the work but cannot close it. The market is starting to price the burden that management still calls discipline.

Friction was built on human limits, then hardened into payroll

Friction is anything that makes it costly to convert a signal into a governed outcome. It is the lag between observing and acting. It is the manual reconciliation that appears when systems disagree. It is the approval chain that exists because permission is fragmented and nobody wants to be the person who acted too early. It is the recurring meeting whose real purpose is to reassemble context that should have traveled with the work. It is the planner, expeditor, scheduler, analyst, coordinator, and manager carrying the distance between what the business knows and what it can actually do. That friction was not created by fools. It was created by limits. Attention is finite. Memory is finite. Coordination capacity is finite. When the work is messy and the penalties for error are high, organizations add layers, procedures, and people whose job is to route context, verify evidence, summarize exceptions, and carry authority from one part of the firm to another. The enterprise learns to call this governance, diligence, and alignment. Each word sounds responsible. That is why the cost can hide for so long. Over time, the temporary answer becomes the operating model. Companies budget for it. Vendors sell into it. Functions grow around it. Careers are built on mastering it. Some of the highest status work in the modern enterprise has been friction management with better vocabulary. The business is not just paying for judgment. It is paying for organizational distance. That is what the repricing is about. Reasoning is getting cheaper. Connectivity is compressing distance. Access is becoming more immediate. The old excuse that separation is merely the unavoidable cost of scale is getting weaker. Separation is becoming more obviously a design choice. Design choices get repriced. Repricing changes employment. Repricing changes capital allocation. Repricing changes which margins deserve a premium and which deserve suspicion. This is why the labor story and the investability story are the same story now. The issue is not whether AI “takes jobs” in the theatrical sense. The issue is whether the enterprise still needs large amounts of human intermediation in the middle just to remain inside acceptable economic bounds. If that is where the margin is coming from, then the margin is under attack. If that is where the business is still hiding delay, then the future is arriving as a repricing of payroll. The old labor story said firms bought talent to solve hard problems. The harder truth is that many firms bought talent to absorb architecture they should already have outgrown.

The original metric was right, but it is no longer enough

The first version of “The Metric That Exposes if Your Transformation Success Is a Lie” put its finger on the right wound. The strongest internal test was not project count, adoption, OEE, headcount optics, or cleaner headline margin. It was percent cost improvement per labor hours consumed, understood broadly enough to include not only direct labor but all the indirect hours tied to execution, quality containment, rework, manual rescheduling, exception handling, supervisory intervention, reporting, expediting, and recovery. That metric still matters because it gets past the vanity of modern change language. It asks what burden actually disappeared. It asks whether the system became easier to run or whether people simply got better tools to keep compensating for the same weakness. It asks whether the operating model still requires the same amount of human effort to produce acceptable results. That is still the right center of gravity. But it is no longer enough by itself. It is not enough because some companies can improve the optics while preserving the dependency. Labor can move off the books into contractors. Headcount can fall while outsourced support rises. Margin can improve because pricing improved, commodity costs moved, a business was sold, or a bad quarter rolled out of the comparison. Inventory can remain heavy while management praises productivity. Cash can lag earnings while the story still sounds clean. A business can appear more modern and still depend on the same burden structure underneath. So the next version of the article has to raise the standard. The internal metric remains percent improvement per total human hours consumed. But that metric only counts if it is paired with proof that the business now closes loops with less carry in the middle. Cash has to follow the gain. Inventory has to stop acting like insurance against indecision. Failure costs have to narrow rather than reappear in new language. Shadow work has to fall. Meetings whose purpose is manual context assembly have to disappear. Roles built mainly around routing, summarizing, and re-litigating exceptions have to contract. Work has to move from titles to boundaries. It has to move toward permission design, evidence discipline, intervention logic, reversal rules, and outcome ownership. That is the future metric. Not a dashboard metric. A verdict metric. A system is not better because it explains more. It is better because it requires less compensation.

The real divide is structural advantage versus friction advantage

This is where the article needs to become more severe. Some businesses create margin through structural advantage. They have product strength, process know-how, cost position, quality credibility, distribution control, switching costs, installed base, geographic strength, or genuine operating superiority that remains valuable when the world gets faster and more connected. Cheap reasoning does not erase those things. In some cases it strengthens them. Other businesses create more of their margin through friction advantage. They are better than peers at living with complexity, routing context, coordinating exceptions, translating between systems, and carrying delay. That can look like excellence. Sometimes it is excellence. But it is fragile because the more of your economics depend on being better at carrying friction, the more exposed you become when the cost of carrying friction starts to fall. That distinction matters because it tells you who is actually doing this well. The firms doing it well will not merely have more software. They will have less dependence on human bridges. Their filings and internal reviews will show margin quality paired with cash quality. Inventory days will ease because the business no longer needs to hide indecision inside buffers. SG&A will stop rising as a tax on coordination. Failure costs such as premium freight, expediting, containment, rework, and recovery labor will narrow rather than migrate. Customer commitments will be made inside the market window more often because fewer decisions need to wait for a room full of people to reassemble the facts. The firms not doing it well will leave a different signature. They will talk about visibility while closure stays weak. They will fund integrations while the same unofficial spreadsheets remain in circulation. They will speak fluently about AI while the same people still spend their afternoons adjusting what the system told them in the morning. They will show cleaner margins in some periods, yet cash will lag, inventory will stay heavy, contractor reliance will deepen, and restructuring or failure charges will keep surfacing because the business is still paying people to recover from itself. That is why the right reading of a company is no longer, “Are they using modern tools?” The better question is, “How much of this business still depends on human intermediation that cheap reasoning is going to mark down?”

The board questions are changing whether boards like it or not

What would happen if every recurring report disappeared tomorrow and only the documents that clearly changed a decision were allowed to survive? How many people would still be doing indispensable work, and how many would be revealed as custodians of a process that exists mainly because the architecture still cannot carry context on its own? If most decisions would not worsen, then the company is paying for narrative production more than decision improvement.

What would happen if your best planner, scheduler, expeditor, or coordinator left tomorrow? Would the business replace that person with another person because the logic still lives in someone’s head, or would it be able to encode the permission boundary, the evidence requirement, the intervention logic, and the reversal rule in a governed system that keeps operating without that single human bridge? If the answer is that people still have to be replaced with people because the architecture cannot hold the model, then payroll is still acting as a substitute for design. Those are not decorative questions. They are the kind of questions that change the room. They diagnose mechanism, not symptoms. They separate a company that is getting easier to run from one that is merely getting better at discussing why it is still hard to run. The old board questions are too soft for the moment. What is our AI strategy is too soft. How many pilots are underway is too soft. How many jobs are exposed is too soft. The more accusing questions are the ones that matter now. How much of our margin depends on middle layers whose main purpose is to intermediate delay? How much of our SG&A is friction payroll rather than durable advantage? How much of our working capital exists because the operating model cannot close loops fast enough to trust itself? How much revenue do we fail to retain because we cannot commit with confidence inside the customer’s window? How much gross margin do we surrender because lateness and uncertainty have become part of the commercial reality? These are not technology questions. They are questions about worth. If new capital does not reduce inferencing burden, it is probably funding a more expensive version of friction.

Why the winners will be obvious before they are universally acknowledged

The next article has to say something embarrassing if it turns out wrong, otherwise it is not carrying enough weight. Here is the prediction. Within the next two annual planning cycles, the difference between firms that have actually reduced burden and firms that have only improved their language will become visible in a repeatable pattern. The winners will show some combination of better margin quality, stronger cash realization, lighter inventory behavior, fewer recovery lines, less contractor substitution for hidden plant and coordination work, and fewer layers of manual carry in the middle. The laggards will keep explaining why the numbers are temporarily distorted, why another system layer is needed, why more visibility is still progress, and why the same recurring meetings remain necessary. If that pattern does not start to separate more clearly, then this argument deserves to be cut back.

The reason the separation should become easier to see is that the old excuses are weakening together. Reasoning is cheaper. Access is less scarce. Connectivity is better. Buyers are getting less patient with explanation without closure. The premium on human intermediation is falling precisely where that intermediation does not produce real boundary ownership, judgment, or controlled action. The firm that still needs large amounts of coordination payroll to keep converting signal into action is going to look more exposed with each cycle, not less. The market does not need perfect theory to price that. It only needs enough repetition to recognize the signature. This is also why the category in the middle matters so much. Not every company will fall cleanly into elite or broken. Some businesses will remain operationally viable but strategically weak. They can still run. They may still generate cash. They may still have excellent people. But they depend too heavily on coordination labor, trapped working capital, retained revenue loss, friction margin, and heroic intervention to deserve long-dated reinvestment without redesign. That is the category leadership avoids naming because it feels like an insult. It is not an insult. It is a capital classification. A plant can still be productive inside a weak business model. A business can have a strong enterprise story built on weak plants whose local inferencing burden is too high to translate strategy into reliable execution. That is why both levels have to be read at once. At the business level the question is whether margin comes from structural advantage or from better tolerance for friction. At the plant level the question is runtime credibility under stress. How long is the time between signal and action? How much depends on tribal knowledge? How many interventions require escalation before they can occur? How much inventory exists as local insurance against slow closure? How much throughput is trapped because the layers above the plant still cannot carry context well enough to support the edge? The firms doing this well will answer those questions with less theater and more evidence. The others will keep confusing motion for redesign.

Why the first true movers become hard to beat

Not all first movers win. Some simply burn capital sooner. Some automate confusion. Some add AI theater to a bad operating model and end up with more burden, not less. Some widen risk because they increase action without increasing permission discipline, evidence continuity, or reversal control. Those companies will be valuable only as cautionary examples. That is the fair counterargument, and it matters. In regulated settings, high-risk environments, and operations where error costs are severe, a certain amount of human intermediation remains rational. Judgment does not disappear. Permission cannot be aband oned. Evidence has to travel with action. A business that cuts too far or encodes too badly can create a faster disaster. There is nothing inevitable about winning merely because one moved early. But that counterargument does not save the laggards. It sharpens the burden of proof.

The companies that move first and true become hard to beat for a deeper reason than software. They stop paying so many people to bridge gaps the architecture should no longer contain. That lowers cost, but cost is only the start. They also reduce decision latency. That improves response speed. Then they release working capital because the business learns to trust its own response time rather than buffering against it. Then they protect more revenue because they can commit inside the customer’s window. Then they protect more gross margin because uncertainty no longer has to be priced into every commercial decision. Then they train a different workforce. Less routing. Less re-litigation. More boundary ownership. More evidence discipline. More controlled autonomy. More people who know how to supervise outcome shaping systems rather than carry the consequences of delay by hand. That is an advantage stack. It compounds. Every loop they close produces clearer evidence. Clearer evidence produces cleaner permission boundaries. Cleaner boundaries reduce escalation. Reduced escalation shortens time to action. Faster action reduces failure cost, trapped inventory, and commercial hesitation. That liberated cash can be turned into more closure. The laggard does the opposite. It takes trapped cash and reinvests it in more intermediation, more support layers, more integration patches, more analytics to explain what still cannot be changed fast enough, and more meetings to carry uncertainty from one title to another. One model compounds control. The other compounds friction. That is why the early true movers become difficult to dislodge. Competitors are not just competing against a better tool. They are competing against a business that has a different cost structure, a different cash profile, a different control architecture, a different labor mix, a different customer response clock, and eventually a different right to capital. By the time the laggard recognizes that the comparison is no longer software against software, it is usually already fighting a company whose operating model has begun to rewrite the commercial terms of competition. The winners will not be the firms with more AI. They will be the firms with less manual carry in the middle.

Permission is where the verdict becomes unavoidable

There is one more reason this article has to be tighter than the earlier draft. It has to be clearer about permission. The constraint is not intelligence. It is permission. Enterprises often talk as if insight were the scarce thing. In many cases it is not. They can already see a great deal. The reason value still leaks is that the path from seeing to doing remains fragmented across titles, systems, and social dependencies. Evidence is scattered. Authority is blurred. Reversal rules are unclear. Nobody wants to act too early, but by the time action is safe

enough for everyone, the value has already started leaking away. So the company buys more visibility, more review, more administrative carry, and more people to mediate the uncertainty. It calls this prudence. It is often permission latency. That is where agency becomes the dividing line. If a system cannot shape an outcome inside a governed boundary, it is not taking on the burden that matters. It may still be useful. It may save time around the edges. It may make meetings shorter or memos cleaner. But it is not removing the real tax in the business if the business still depends on the same people, in the same rooms, to carry the logic, the evidence, and the authority required to act. This is why the future labor map is not really about titles. Titles can linger. Boundaries do not wait. The work that remains valuable is the work tied to judgment, evidence, boundary control, intervention design, and accountable outcome ownership. The work under pressure is the work whose primary purpose is to route context, translate between systems, summarize exceptions, carry approvals, and absorb organizational distance by hand. That is not a moral verdict on the people in those roles. It is a design verdict on the business that still needs so many of them. The article has to say that without flinching. Too much of the modern enterprise has become a payroll structure built to compensate for weak closure. That was survivable when reasoning was expensive. It is much harder to defend when reasoning is cheap and the market can begin to ask, with increasing bluntness, why so much of the company still exists to carry the distance between knowing and doing.

The final classification will be harsher than most management teams want

Some businesses and plants will remain fully investable. They will reduce inferencing burden, shorten decision chains, clarify authority, improve evidence continuity, release trapped working capital, protect revenue, and convert new capital into real closure. Their complexity may still be real, but the burden of carrying that complexity will not keep rising in proportion. They will deserve reinvestment because each dollar placed into them improves control rather than merely financing more intermediation. Some businesses will remain investable, but only with redesign. Their recoverable pools will still be large enough, and the path to better closure still plausible enough, that capital can earn a strong return if leadership is willing to stop funding theater and start reducing burden. Some will remain operationally viable but strategically weak. These are the companies that can still ship, still explain, still defend the quarter, and still keep good people working very hard, yet the underlying economics depend too heavily on friction payroll, trapped cash, retained revenue leakage, and repeated recovery labor to justify long-dated confidence.

And some are already drifting toward uninvestable, even if nobody in the executive review wants to use that word. They will remain in motion. Systems will rise. Meetings will rise. Layers will rise. Support labor will rise. Visibility will rise. Closure will not. The enterprise will become better at seeing what it still cannot change in time. At that point the question is no longer how to optimize the asset. The question is whether leadership will name the verdict before the market does it for them. That is where the revised article has to end. Not in advice. Not in a playbook. In inevitability. The file on the table at 9:00 a.m. already knows which category the business is in. So does the customer who needed the answer by Friday and found it elsewhere two weeks later. So does the planner whose work keeps getting praised precisely because the architecture still cannot hold the model. So does the CFO who sees margin improve on paper while cash lags and inventory refuses to behave. So does the board, even when it keeps asking softer questions because the harder ones threaten the story the company has been telling about itself. The future will not separate firms by who talked best about AI. It will separate them by who reduced the burden in the middle, who shortened the distance between signal and action, who encoded permission with discipline, who carried evidence with the workflow, and who stopped paying so many people to compensate for what the system should already know how to do. Cheap reasoning will not kill the enterprise. It will expose it.

References

This piece draws first on Michael Carroll’s own work on decision latency, permission as the architecture of risk, burden transfer, friction repricing, agency as the ability to shape an outcome, and the one degree reality in which connectivity and access compress separation and force a harsher capital verdict. It also builds directly on the supplied article set, especially The Metric That Exposes if Your Transformation Success is a Lie for the labor burden test and the corrected reading of margin, cash, inventory, and failure cost, How to Know the Businesses and Plants That Still Have an Investable Future for inferencing burden, structural versus friction advantage, and the classification of investable versus merely operable assets, and What Jobs Will Be Left After AI? This Article Actually Answers That for the move from titles to boundaries and the claim that friction is payroll. The broader ballast comes from Ronald Coase’s The Nature of the Firm in 1937 on coordination cost, Herbert Simon’s Administrative Behavior in 1947 on bounded rationality, W. Edwards Deming’s Out of the Crisis in 1986 on system quality and rework, Jay Galbraith’s Designing Complex Organizations in 1973 on information processing load, Dixit and Pindyck’s Investment Under Uncertainty in 1994 on timing and option value, and Judea Pearl’s Causality in 2000 and The Book of Why in 2018 on the difference between observation and intervention, because the argument here is not that companies need more explanation, but that they are being judged ever more directly on whether they can convert explanation into governed action before time turns against them.