Good Enough Is Now Too Expensive

Michael Carroll | The One-Degree Dispatch | Critical Decision Infrastructure

Good Enough Is Now Too Expensive Why the way executives really decide became a hidden tax on COOs, and why modern enterprises need Critical Decision Infrastructure before AI, peers, or productivity work can matter. By Michael Carroll Founder, The One-Degree Dispatch | Research Fellow | Board Advisor | Industrial Transformation Leader

Lead image. The title scene places the article inside the executive room where local discipline can still produce enterprise delay.

The operating review does not look broken from the outside. The packet is prepared. The numbers are current. The leaders at the table know their functions. Finance has the variance. Operations has the recovery plan. IT has the system constraint. HR has the staffing issue. Commercial has the customer pressure. Everyone is doing the responsible thing from the seat they occupy. That is what makes the pattern so costly. A late shipment becomes a review. A quality miss becomes a containment plan. A productivity gap becomes a weekly update. A digital pilot that did not scale becomes another governance discussion. The company does not call this waste. It calls it discipline. It calls it alignment. It calls it risk management. Each word has a reason to exist. The problem is what happens when the organization becomes better at managing consequences than changing the conditions that keep producing them. The COO feels that bill first. This is not a story about bad executives. It is a story about competent people making the best workable decision they can inside systems that were never built for the speed, fragmentation, and uncertainty now running through the enterprise. Herbert Simon gave that behavior a name. He called it satisficing. The word Michael Carroll | The One-Degree Dispatch

will not help in most executive rooms, but the behavior is everywhere. People do not optimize against every possible future. They search until they find an answer that is good enough to move, safe enough to defend, and acceptable enough to survive the organization. The old management story said executives should maximize. The real enterprise never worked that way. It made decisions with partial information, limited time, uneven authority, political exposure, functional incentives, and incomplete evidence. Simon saw that clearly. Later, Daniel Kahneman and Amos Tversky showed why judgment bends further under uncertainty, even when intelligent people are trying to be careful. Together, their work explains something every COO already knows. The company does not usually fail because no one cares. It fails because local decisions that make sense in isolation can still produce a system that consumes capacity, slows action, and misses the outcome.

Good enough becomes expensive when the system keeps producing the same bill

The question is no longer whether executives can find the perfect decision. They cannot. The question is whether the enterprise has built the conditions that make a good enough decision good enough for the whole company, not merely for the function, the meeting, or the quarter. That is where the COO’s work has changed. The COO is no longer only the executive responsible for operations. The COO is increasingly responsible for the conditions under which the enterprise turns intent into outcome. That requires a different kind of infrastructure.

The company did not fail to decide. It decided inside fragments

A fragmented enterprise can look mature. It has sharper functions, deeper systems, more specialized leaders, clearer metrics, more governance, and more expert review. Each piece becomes stronger on its own terms. Finance gets more disciplined. IT gets more protective of architecture. HR gets more alert to adoption and workload. Commercial gets more responsive to customer commitments. Operations gets more precise about what can actually be run. Then the pieces meet. The CFO may be right about capital discipline. The CIO may be right about integration risk. HR may be right about the burden on people. Commercial may be right about market timing. Operations may be right about feasibility. None of those local truths guarantee enterprise coherence. They can all be true, and the company can still move too slowly, spend too much capacity on workarounds, and miss the condition required for the objective.

Michael Carroll | The One-Degree Dispatch

Supporting image. The handoff is where a locally reasonable decision can become an enterprise delay.

This is where Simon’s insight becomes practical for COOs. In real organizations, people search within boundaries. They do not search the whole landscape. They search the part they can see, the part they can defend, the part their role allows them to influence. The decision that emerges is not the best possible decision. It is the decision that satisfies enough local constraints to move through the system. That works when the environment is stable, the functions are tightly coupled by habit, and the cost of delay is small. It works less well when the enterprise is operating across global supply chains, constrained labor, volatile demand, fragile technology stacks, and AI programs whose value depends on data, permissions, process, and human trust all working together. Under those conditions, the old model does not simply slow the company down. It teaches the company to treat the symptoms of weak control as if they were the work itself. A recurring escalation is not necessarily a sign of leadership discipline. It may be evidence that authority sits too far from the signal. A weekly recovery meeting is not necessarily proof of accountability. It may be proof that the system keeps creating the same consequence. A dashboard is not necessarily evidence. It may only be a record of what has already happened. A governance process is not necessarily control. It may be a mitigant that consumes capacity after the operating condition has already failed. What would have to be true for this outcome to keep repeating. That question changes the altitude of the conversation. It moves the COO away from chasing the late order, the missed target, the failed pilot, or the repeated exception as isolated events. It forces the leadership team to ask what landscape they are operating on, what levers they can actually pull, what controls they have placed there, and what conditions those controls are creating. If the conditions do not support the objective, the next review will only be a more polished version of the last one. Michael Carroll | The One-Degree Dispatch

The figure shows why recurring mitigants are not neutral. They consume the leadership and operating capacity needed to improve controls.

Kahneman made the risk harder to ignore

Simon showed that people decide under limits. Kahneman and Tversky showed that judgment under those limits is not merely incomplete. It is patterned. People anchor on early information. They overweigh what is vivid or recent. They rely on examples that come easily to mind. They protect against losses more heavily than they value equal gains. They judge from reference points, not from a clean economic field. Executives are not exempt from this. They may be more experienced, but they are also carrying more consequence. A COO who has been burned by a failed rollout will read risk differently the next time. A CFO who has seen benefits fail to materialize will demand a harder case. A CIO who has inherited technical debt will see fragility before opportunity. HR will remember the last change program that exhausted the organization. These are not irrational reactions. They are lived evidence. The question is whether the evidence is still describing the current system or anchoring the team to the last scar. The COO has to manage that tension without pretending judgment can be purified. Judgment is always carried by people with memory, role, incentive, and exposure. That is why the answer cannot be more data alone. More data can even make the problem worse when it gives every function more material to defend its own local conclusion. Data does not become evidence until there is a question. The question determines what the observation is evidence for or against. Without that discipline, the enterprise confuses reporting with learning.

More data does not create agency unless it changes what the organization can do

Consider the familiar AI pilot that succeeds in one part of the business and then stalls. The common explanation is adoption. Sometimes that is true. But the deeper question may be whether the pilot ever changed a control. Did it move authority closer to the work? Did it improve evidence quality? Did it reduce the capacity spent on exceptions? Did it clarify who could act and when? Did it change the operating condition that produced the consequence in the first place? If the answer is no, then the pilot may have improved a task while leaving the enterprise burden intact. Michael Carroll | The One-Degree Dispatch

That is where Kahneman’s work becomes more than a lesson about bias. It becomes a warning about the enterprise itself. A fragmented company gives every function its own reference point. The finance reference point is the forecast. The operations reference point is what the system can run. The technology reference point is risk to the architecture. The workforce reference point is what people can absorb. The commercial reference point is the customer commitment. Each reference point is legitimate. The COO’s burden is to prevent those reference points from becoming separate realities. The strongest counterargument deserves respect. In many operating settings, good enough decisions save the business. The line does not wait for an academic debate. The customer does not wait for a perfect causal model. The quarter does not wait for flawless evidence. Experienced operators often make fast calls that outperform slow analysis, especially when the work is familiar, the feedback is immediate, and the cost of waiting is clear. That counterargument does not weaken the case. It defines it. The goal is not to eliminate good enough decisions. The goal is to make the system better at defining what good enough must mean. In a strong operating system, good enough is bounded by clear authority, trusted evidence, defined controls, known thresholds, and fast learning. In a weak operating system, good enough is negotiated through meetings, personalities, risk transfer, and the need to keep the business moving. The difference is control.

COOs do not own every cause, but they own the operating bill

The COO’s problem is that consequences rarely arrive with clean ownership. A customer miss may originate in demand planning, production scheduling, supplier performance, labor availability, transportation capacity, and commercial promise rules. A margin issue may pass through mix, yield, overtime, scrap, inventory, logistics, and pricing discipline. A quality problem may include equipment condition, standard work, training, supplier inputs, inspection logic, and management cadence. Each function can explain its part. The COO has to make the enterprise perform as one system.

Michael Carroll | The One-Degree Dispatch

Supporting image. The COO sees the operating bill because recurring consequences appear as load on the system.

That is why the COO role is moving from operational supervision toward operating control. Not control as command. Control as the set of mechanisms that create the conditions required for the objective. Decision rights are a control. Standards are controls. Measures can be controls if they change action before the consequence appears. Cadence can be a control if it improves learning rather than merely adding review. Software architecture can be a control if it changes what the system can sense, authorize, and do. Governance can be a control if it moves judgment closer to the work and clarifies accountability. It is only a mitigant if it adds review after the system has already produced the consequence. This distinction matters because mitigants are expensive. They consume leadership time, operating focus, trust, meeting capacity, analytical effort, and attention that could have been used to improve the controls. A quality escape triggers containment. A slow decision triggers escalation. A missed productivity target triggers review. A failed pilot triggers more oversight. Some of those actions are necessary once the consequence exists. Necessity does not make them causal control. The hidden cost center is not always the missed target. It is the permanent capacity assigned to managing the recurring miss. When the same issue keeps returning, the organization does not merely pay for the event. It pays for the meetings, reconciliations, reports, workarounds, approvals, explanations, and credibility loss that surround it. The COO sees that cost because it appears as load on the system. People are busy, but the business is not improving at the rate the activity implies. A practical test is simple enough to use in an executive meeting. When a consequence appears again, can the team say what condition produced it, what lever can change that condition, what control should be placed on the landscape, and what outcome should improve if the control works? If the team cannot answer that without retreating into functional explanations, it is not yet working on the system. It is managing the bill. Michael Carroll | The One-Degree Dispatch

Another test cuts into permission. Where does the signal first appear, where does authority to act sit, and how much capacity is consumed between the two? If the signal appears near the work but authority sits three meetings away, the company has designed delay into the system. If every exception needs escalation, the escalation process has become the real operating model. If the same dashboard shows the same misses every month, the dashboard is not learning. It is testifying.

A dashboard is not control unless it changes what the system can do

This is why the COO cannot solve the problem through personal force alone. Heroic intervention can rescue an outcome, but it cannot be the operating model. The more the enterprise depends on the COO’s ability to break through friction, the more it proves the infrastructure is weak. A strong COO does not simply chase harder. A strong COO improves the conditions that reduce the need to chase.

Critical Decision Infrastructure is the COO’s control language

Critical Decision Infrastructure is not another term for governance. It is not a committee, dashboard, operating review, or AI platform by itself. CDI is the structure that helps leaders understand the consequence landscape, identify the levers that matter, apply the right controls, and create the operating conditions that make the objective more likely to become true. For a COO, that definition has to stay close to work. It begins with the objective. What condition is the company trying to make true? More throughput. Better reliability. Faster order fulfillment. Lower working capital without service damage. Safer work. Higher quality. Better labor productivity. Faster benefit realization from AI. The objective matters because it creates the question. The question determines what data can become evidence. The evidence tests what is true, what is assumed, and what must be changed. The operating chain is not complicated, but companies break it constantly. Objective creates question. Question creates hypothesis. Observation becomes evidence only when it answers the question. Evidence supports a causal model. The causal model identifies what would have happened under different conditions. The intervention changes the system. Controls create new operating conditions. Conditions produce outcomes. Outcomes feed learning back into the next decision. The enterprise does not need to recite that chain. It needs to operate as if it were true.

Michael Carroll | The One-Degree Dispatch

CDI matters because it links the objective to the controls and conditions that make the outcome more likely to become true.

That is what CDI does when it is real. It prevents the company from treating consequences as causes. It prevents mitigants from pretending to be controls. It prevents reports from pretending to be learning. It prevents governance from pretending to be agency. It gives the COO a way to connect strategy, finance, technology, workforce, and operations without reducing the conversation to slogans about alignment. CDI also gives the COO better language with peers across the executive team. With the CFO, CDI connects operating conditions to financial outcomes. It shows whether a margin problem is being managed through explanation or changed through control. With the CIO, CDI connects architecture, data, permissions, identity, and system writeback to the work the company needs to perform. With HR, CDI connects role clarity, capability, trust, and adoption to the operating condition rather than blaming people for system friction. With the CEO and board, CDI changes the conversation from effort to control. The peer element matters because every company normalizes its own workarounds. Over time, the exception report becomes normal. The escalation meeting becomes normal. The manual reconciliation becomes normal. The approval chain becomes normal. People stop seeing these as capacity drains because the organization has built routines around them. Serious peers help break that normalization. They ask why the same consequence still needs attention. They ask whether the control is placed on the right landscape. They ask whether AI is changing the work or merely speeding up the report.

Michael Carroll | The One-Degree Dispatch

The figure shows why peers and CDI belong together. The peer room tests the explanation, and CDI turns it into control.

This is not networking. It is external discipline for internal clarity. A good peer room does not flatter the COO. It sharpens the COO’s questions. It gives the role a way to compare patterns, test assumptions, identify better controls, and bring a stronger language back into the company. In a fragmented enterprise, the COO needs peers not because the work is lonely, though it often is, but because no single company sees its own operating system cleanly enough from inside the room.

AI will punish weak infrastructure before it rewards ambition

AI raises the stakes because it can make both good and bad systems faster. A company with weak decision rights can automate confusion. A company with poor evidence standards can accelerate unsupported recommendations. A company with unclear authority can create more exceptions at machine speed. A company that confuses mitigants with controls can use AI to process the consequences faster while leaving the conditions untouched.

Michael Carroll | The One-Degree Dispatch

Supporting image. AI becomes material only when it changes a control, improves a condition, and preserves accountability.

The attractive version of the AI story says the technology will reduce friction, improve decisions, and increase productivity. That can be true. But the claim is incomplete unless the enterprise can say where the control changes. Does AI improve the question? Does it convert observation into evidence? Does it identify the condition behind the recurring consequence? Does it recommend an intervention? Does it place authority closer to the work? Does it preserve human judgment where accountability belongs? Does it learn from the outcome? If not, it may be useful, but it is not yet changing the operating system. By the end of 2028, the companies that get durable operating value from AI will not be the ones with the most pilots. They will be the ones that can show where AI changed a control, improved an operating condition, and reduced the capacity spent managing recurring consequences. That prediction is falsifiable. If companies with broad AI deployment but weak authority, weak evidence standards, and slow permission structures consistently outperform companies with fewer AI deployments but stronger controls and faster learning loops, then this argument is weaker than it appears. The observed pattern today is less settled than the market language suggests. Many companies are moving quickly. Many are also struggling to connect AI activity to earnings-relevant operating change. That is an observation. The inference is that AI value depends heavily on whether the operating landscape has the right controls. The projection is that CDI will become one of the dividing lines between companies that scale AI into performance and companies that scale AI into more sophisticated activity.

AI does not remove the need for control. It raises the price of weak control

Michael Carroll | The One-Degree Dispatch

This is where the COO’s role becomes decisive. The COO should not be the executive who merely receives AI from somewhere else in the enterprise. The COO has to ask where AI sits in the operating chain. Is it sensing conditions? Is it improving evidence? Is it supporting the causal model? Is it recommending interventions? Is it authorizing action within clear boundaries? Is it learning from outcomes? Is it releasing capacity from recurring mitigants? The question is not whether AI is present. The question is whether it helps the company make the desired condition true. If it cannot do that, it may still be useful. It may save time. It may summarize. It may route. It may reconcile. Those benefits should not be dismissed. But a COO has to separate useful task improvement from operating control. The former may improve efficiency. The latter changes performance.

The next COO standard will be judged by what no longer repeats

The best COOs have always known that operations is not the art of explaining yesterday. It is the work of creating conditions that make tomorrow’s result more likely. What has changed is the size of the condition set. The modern COO has to deal with finance, technology, people, data, AI, supply chain, commercial commitments, risk, and execution as one operating landscape. The old function-by-function answer is too slow. The old hierarchy is too far from the signal. The old governance routines can consume the very capacity needed to improve the system. That does not mean the COO should reject governance, dashboards, reviews, or escalation. It means those tools have to be judged by a harder standard. Do they improve the control system? Do they move judgment closer to the work? Do they reduce latency? Do they improve evidence? Do they clarify authority? Do they make the desired condition more likely? If they do not, they may only be well-dressed mitigants. The next COO standard will not be measured only by how well leaders respond to consequences. It will be measured by how many recurring consequences the system stops producing. That standard is less theatrical than heroics. It is also more demanding. It requires the COO to see the landscape, find the levers, apply controls, and release capacity from work the organization should not need to keep doing. This is why peers and CDI belong together. CDI gives the COO the language of control. Peers keep the language honest. The enterprise is too good at defending its own habits. A serious peer room can ask whether the company is changing conditions or only managing the bill. It can test whether the apparent cause is really a consequence. It can force the distinction between the measure, the mitigant, and the control. It can help COOs see their work not as isolated operating problems, but as recurring patterns across companies facing the same structural pressure. The deeper lesson from Simon, Kahneman, and Tversky is not that executives are flawed in some special way. It is that human judgment has limits, and organizations often amplify those limits through fragmentation, delay, and local incentives. The answer is not the fantasy of perfect optimization. It is better control over the conditions in which judgment has to act. Companies will keep making good enough decisions. They have to. The work will not stop until certainty arrives. The question is whether good enough is defined by local safety, meeting survival, and familiar mitigants, or by evidence, authority, controls, and learning strong enough to improve the next outcome. The COO is the executive most likely to know the difference because the COO owns the operating bill when the system gets it wrong. That is why the role now sits at the center of the next enterprise standard. Not because operations is everything, but because every strategy eventually has to become an operating condition before it can become a result. If the same consequence keeps returning, the system is still telling the truth.

Michael Carroll | The One-Degree Dispatch

References

This article draws on Herbert Simon’s work on bounded rationality and satisficing, including his 1978 Nobel Prize lecture on decision-making in business firms, Daniel Kahneman’s 2002 Nobel recognition for integrating psychological research into economic decision-making under uncertainty, Amos Tversky and Daniel Kahneman’s 1974 Science article on heuristics and biases, Kahneman and Tversky’s 1979 work on prospect theory, Cyert and March’s behavioral theory of the firm, and Jay Galbraith’s informationprocessing view of organization design. These sources matter because they explain why perfect optimization was never a realistic description of enterprise behavior, why judgment bends under uncertainty, why organizations search locally and resolve conflict through process, and why greater uncertainty requires better information processing and control design rather than more consequence review. The article also draws on Michael Carroll’s prior work on Critical Decision Infrastructure, human agency, causal evidence, controls, mitigants, operating conditions, and the claim that data does not become evidence until there is a question.

Michael Carroll | The One-Degree Dispatch