The Synchronization Constraint


Why Fast Teams Still Lose

A company can correctly diagnose the bottleneck, own the constraint, and build real execution authority, and still lose. Not because the diagnosis was wrong. Because the parts of the organization that needed to move together didn’t.

The Strategy Isn’t Wrong. The Clock Is. argued that industries operate at different rates of change, and that value migrates to whichever constraint is binding at a given moment. This essay asks a narrower question, one that only shows up after a company has already gotten the market-level diagnosis right: why do organizations with the correct strategy, the correct constraint, and even the correct authority still fail to capture the value they correctly identified?

The answer isn’t speed. It’s coupling.

Why Speed Alone Isn’t the Problem

If engineering ships ten times faster than HR, nobody notices. The two functions barely touch. Speed mismatches only matter when the functions involved are tightly coupled, when one cannot act without the other’s output, sign off, or clearance. A company can run five, ten, twenty different clocks simultaneously without consequence, as long as most of those clocks operate independently.

The failure mode shows up specifically where two interdependent functions must clear each other on every cycle. The constraint isn’t the slowest function in the building. It’s the slowest dependency that every critical cycle has to cross.

Up to a point, faster engineering improves throughput. Beyond a certain synchronization gap, additional speed reduces throughput, because every new release creates more unresolved work for the coupled function downstream, more risk inventory, more exceptions stuck in a queue that was never built for this volume. That’s a testable claim, not just a plausible one. Track deployment frequency against governance approval frequency against channel update frequency in a real organization, and watch what happens to throughput as the variance between them widens.

Ford Model e and the Channel Clock

In 2022, Jim Farley split Ford into separate units, creating Model e to run the EV business at software speed while legacy combustion stayed under Ford Blue. The diagnosis was right. Competing with direct-to-consumer software platforms meant transparent pricing, digital purchasing, and real charging infrastructure.

Model e’s product and software teams moved fast. To match that on the commercial side, Ford introduced the Model e Certified dealer program that September, asking franchised dealers to invest up to $1.2 million for fast chargers, non-negotiable pricing, and digital sales workflows.

Here’s the coupling. Ford’s software team didn’t need dealers to ship code. But it did need every one of them to clear a state franchise law before the new pricing model could go live, and that clearance is the dependency this essay is about. The dealer channel clock was governed by decades of statute Ford could not rewrite on its own. Dealers in at least six states filed suits and administrative challenges arguing the mandates violated dealer-protection law. Because that clock ran on court schedules and statutory notice periods, Ford couldn’t force the model through. By November 2023 it rolled back the requirements. By July 2024 it scrapped the certified dealer program entirely.

Worth being honest about the full picture. EV demand was cooling industry-wide during this same window, and Ford pulled back roughly $12 billion in EV spending for reasons well beyond dealer friction. Ford doesn’t prove the law by itself. What it does is illustrate the mechanism cleanly, a fast function coupled to a slow one it could not bypass, with the slow one setting the outcome regardless of how good the fast one was.

GE Digital and the Coupled Sale

The same pattern shows up again in GE Digital’s Predix platform through the 2010s, through a sales dependency instead of a legal one, and the coupling here is almost textbook. GE correctly diagnosed that industrial equipment, turbines, jet engines, locomotives, was shifting toward software-driven predictive maintenance, and spent billions building the capability with real engineering autonomy behind it.

Engineering shipped on two-week cycles. But every dollar of that software had to pass through a sales force compensated on multi-million-dollar, multi-year capital deals. Engineering could not monetize its output without that function clearing it first. Asking a rep whose commission depended on a $100 million turbine contract to sell a $50,000 annual subscription wasn’t a parallel problem sitting next to engineering. It was the coupling point engineering’s output had to pass through to become revenue.

Worth being honest here too. GE Digital’s wind down was tangled, not clean. Business units that had already built competing tools resisted adoption, the strategy spread across too many verticals at once, and the platform had real technical shortcomings of its own. The sales coupling wasn’t the sole cause. It was a genuine, structural brake.

What This Costs

Large organizations run on more than a handful of clocks, product, governance, commercial channels, procurement, finance, regulation, and which ones matter varies by industry. The specific clocks aren’t the point. The coupling between them is, and every one of those coupling points is really an interface built for one clock speed and now being asked to serve another.

What asynchronous coupling costs rarely shows up on a quarterly report. It shows up as risk inventory nobody signed off building, capital held against decisions stuck in a queue, learning that arrives too late to change the next cycle. As execution compresses toward real time, advantage migrates away from the teams that move fastest and toward the organizations that synchronize their critical dependencies most effectively. The scarce capability is no longer building faster teams. It’s synchronizing the dependencies between them.

The Constraint Capture Matrix


Finding a bottleneck is not the same thing as capturing it.

Every technology cycle creates the same illusion. Leaders locate the friction point in their industry, buy software to address it, and expect market share to follow. The historical pattern says otherwise. Some companies convert that diagnosis into a multi year moat while others just automate their existing limitations.

In The Strategy Isn’t Wrong. The Clock Is., I argued value migrates to whichever constraint is binding at a given moment. In The Scarce Thing, we saw what happens when that constraint becomes permission. Diagnosing a bottleneck tells you where value is trapped. It tells you nothing about whether your organization can actually claim it.

Two independent variables decide that. Does the organization control the constraint, and has it built the authority to act on that control? Advantage doesn’t come from finding the next constraint, your competitors will see it too eventually. It comes from controlling the constraint and redesigning the permission structure to act on it. The final scarce resource is authorized action.

Control vs. Authority

Constraint control is the control surface. A company rarely owns a bottleneck outright, regulators own rules, payers own reimbursement, markets own capital costs. What a company can build is the specific point where it can influence the outcome without controlling the entire system. High control means the deciding variables sit inside your software, your balance sheet, your data, your infrastructure. Low control means clearing the bottleneck requires independent third parties, regulators, licensing boards, platform monopolies, to move together.

Execution authority is the permission structure, an organizational choice to delegate action without a human calendar queue. It requires runtime verification, pre audited risk bounds, and clearing the liability traps that make managers hoard sign off. Most AI transformations don’t stall for lack of use cases. They build intelligence systems inside organizations still running on human speed permission.

Four positions follow from these two axes. Quadrant 1, Default Capture, high control and high authority. Quadrant 2, the Sunk Cost Engine, high control and low authority. Quadrant 3, the Fragile Velocity Trap, low control and high authority. Quadrant 4, the Dependency Trap, low control and low authority. Companies move between them along recognizable paths: control without new authority takes you from Quadrant 4 to 2, authority without control takes you from 4 to 3, securing a control surface over your dependencies moves you from 3 to 1, and replacing manual queues with real permission architecture moves you from 2 to 1.

This matters now because AI compresses the cost of insight faster than organizations can redesign around it. The bottleneck is moving from generating answers to authorizing actions. The companies that win won’t necessarily have the best models. They’ll have the shortest path between machine insight and business execution.

The Four Quadrants

Quadrant 1 is the target state. Amazon Logistics is canonical. Delivery reliability, not inventory selection, became the real promise to the customer, and third party carriers were a bottleneck Amazon couldn’t dictate. By building its own fulfillment centers and last mile fleet, Amazon built the control surface itself, and every routing improvement compounded into its moat. Credit unions like FORUM, Centris, and Del-One did the same thing in lending. Underwriting guidelines and software sat inside their own walls, so they controlled the constraint, and granting execution authority to pre audited, bounded transactions turned that into a 70% jump in processing capacity, the case covered in The Scarce Thing.

Quadrant 2 is where most enterprise technology budgets die. In corporate procurement, AI can ingest contracts and draft renegotiated terms in seconds, fully within the company’s control. CFOs still refuse to delegate execution, a rational response until machine agency liability is settled in court. The system drafts faster. A human still signs everything.

Quadrant 3 is speed without ownership. Marcus by Goldman Sachs built genuine machine speed underwriting, real authority by any definition, on top of an economic foundation it didn’t control. A new accounting rule, CECL, forced far more aggressive loss reserving just as rates made the loan book more expensive to carry. Worth repeating the honest caveat here rather than dropping it: some of the loss likely traces to underwriting quality itself, Goldman’s card loss rate ran above peer issuers, which is an execution problem, not a control problem. Both can be true. Quadrant 3 doesn’t require a company to be blameless, only that part of the outcome trace to a constraint it never controlled, and here it does.

Pear Therapeutics belongs here too. Its software tracked and adapted treatment for substance use disorder without clinician micromanagement, real authority over the clinical action itself. What it never had was control over the constraint that decided its fate, CMS and commercial payers never assigned a reimbursement code that made the product viable at scale. Pear cleared the hard regulatory hurdle and still had no leverage over the variable that actually mattered.

Quadrant 4 is the default posture of most regulated industries, because it takes no decision to end up there. Traditional prior authorization is the clearest case. AMA surveys consistently find the overwhelming majority of physicians say it delays necessary care, and most of the industry still runs on fax and phone holds on both sides. No one owns the reimbursement rules, and no one built software to act quickly within them either. It’s the quadrant every other example in this chapter started from.

Cohere Health vs. Olive AI

Both companies attacked prior authorization from Quadrant 2. Olive AI raised over $900 million to a $4 billion valuation using robotic process automation to fill out payer web forms faster. It automated the paperwork, not the authority, hospital staff still verified and submitted every request, because Olive had no integration into payer decision engines. It shut down in October 2023, selling its prior authorization unit to Humata Health. In fairness, overexpansion and a burn rate above $100 million a year played a real part too, not every dollar of that failure is a pure Quadrant 2 story. But the core pattern holds.

Cohere Health partnered directly with Humana in 2021 to embed the payer’s own coverage policies into its platform at the point of care. Median approval time on musculoskeletal requests dropped to zero minutes, with 89% approved for immediate scheduling. The partnership is now nationwide, and Cohere reports real time approval on up to 85% of documented submissions. Olive optimized Quadrant 2. Cohere built a new control surface and moved to Quadrant 1.

Beyond AI

This move predates artificial intelligence. Apple built the App Store’s review and payment architecture into an unassailable control surface over mobile distribution. NVIDIA built CUDA into a software layer that binds developers to its silicon. Tesla built the Supercharger network to remove a constraint utilities weren’t going to solve for it. Medtronic used FDA’s Predetermined Change Control Plan to pre negotiate future updates to its LINQ II monitor, turning a recurring regulatory toll into its own iteration schedule. Every one of these firms treated the binding constraint as something to be redesigned, not accepted.

The Guardrail and the Audit

Control and authority are necessary, not sufficient. A company can execute flawlessly inside a domain it fully controls and still fail if it misdiagnosed where scarcity actually lives. The matrix starts only after that diagnosis is right.

Before allocating capital, ask three things. Which quadrant does this initiative occupy now? Do you control the constraint, or are you building a Fragile Velocity Trap? And what organizational cover is keeping you in Quadrant 2, what would it take to authorize the bounded action instead of routing the whole system through a human queue?

Diagnosing the bottleneck tells you where value is migrating. The matrix tells you whether you’re actually equipped to capture it.

Intelligence Is Cheap. Permission Still Has to Be Built by Hand.


Toyota did not win the manufacturing wars of the 1980s because it had a better factory. American plants ran comparable equipment, comparable tolerances, often the same suppliers. Toyota won because it redesigned who had permission to act.

In the 1970s, Toyota gave line workers something most manufacturers would have considered reckless: the authority to stop the entire production line. The worker who pulled the andon cord wasn’t the most senior person in the building. They weren’t in a meeting with the plant manager. They were usually just the person standing closest to the defect, the one with the least formal power and the most immediate information. The andon cord was never a productivity tool. It was an authority architecture, a decision about who gets to act on what they see, without waiting for someone above them to see it too.

American manufacturers spent the better part of a decade copying the visible parts, the kanban cards, the quality circles, and mostly failed, because what they were copying wasn’t the actual advantage. The advantage was the redesigned permission underneath it, and permission doesn’t show up on a factory tour.

A Second Proof, Twenty Years Later

I’ve written before about Desktop Underwriter, the automated mortgage underwriting system Fannie Mae shipped in 1995. The value didn’t come from software that could evaluate a loan file faster than a human. Every competitor could eventually buy comparable software. The value came from what Fannie Mae attached to the system’s output: a waiver, relief from having to re-verify certain judgments the system had already made, provided a lender’s own data and documentation held up. That’s not automation. That’s an institution redesigning who could decide, who carried the risk when the decision was wrong, and how exceptions got handled, around a machine’s output.

Same mechanism, different industry, twenty years apart. The technology was necessary in both cases. It was never what got captured. What got captured was the permission architecture built around it, and I’ve called that capture Default Capture before: the winner is never whoever owns the technology, it’s whoever owns the constraint the technology creates downstream of itself.

AI is about to run the same test a third time.

The Question That Actually Matters

Most of the AI conversation happening in boardrooms right now is stuck on one question: can the model reason well enough to be trusted?

That’s the easy question, and the mortgage industry answered its version of it in 1995. The harder question, the one that actually determines whether an enterprise can act on what its systems produce, is different:

Can the organization allow the model to act?

Every technology transition creates a new abundance somewhere. The winners are never the organizations with the most of the newly abundant thing. They’re the ones that redesign permission around whatever became the constraint instead, before the market redesigns it for them, on someone else’s timeline.

That’s not a technology question. It’s the same question Toyota answered on a factory floor and Fannie Mae answered in a rulebook. Most enterprises haven’t answered it at all. They’ve bought the equivalent of the andon cord and left it bolted to the wall, unconnected to anything.

The Pilot Trap

Walk into almost any large enterprise right now and you’ll find the same three things: an enterprise AI license, a dozen point-solution pilots, and a Center of Excellence generating slide decks about theoretical time savings.

The demos are genuinely good. That was never the problem. The problem is that the company is putting a new engine into an old transmission. Every output still moves through the same approval chain built when the model needed supervision to be trustworthy. Every exception still follows the same escalation path designed for a system that used to be wrong a lot more often than it is now.

The model got dramatically better. The org chart didn’t get the memo. That gap is where the money goes to die, not in model cost, in the friction of an organization still authorizing decisions the way it did when authorization was the only safety mechanism available.

I don’t think this is a technology adoption problem. It’s an Organizational Rewiring Latency problem, and it’s a particularly nasty one, because unlike most of the shifts I’ve written about, this one doesn’t announce itself as a crisis. Nothing breaks. The pilots keep running. The demos keep landing well in the quarterly review. The company just quietly stays exactly as slow as it was before, with a much more expensive engine attached to the front of it.

The Blast Radius Problem

I’ve written before about the Law of Migrating Scarcity: when technology makes something abundant, value doesn’t disappear, it moves to whatever the abundance can’t dissolve. For decades, enterprise intelligence was that scarce resource, so companies built permission systems that assumed it would stay that way, slow, expensive, routed through whoever in the hierarchy had the most of it. Permission itself was never the scarce thing. It didn’t have to be. It only had to keep pace with a world where a decision moved as fast as the human hierarchy that had to bless it.

That world is ending. Models are commoditizing on schedule, the same way the underlying intelligence is, and the permission system built around its old scarcity is what’s left standing as the constraint. Permission isn’t becoming scarce out of nowhere. It’s being exposed as the bottleneck it was always going to become, the moment the thing it was rationing stopped being rare.

The advantage now belongs to whoever builds the clearest architecture for what a system is allowed to do without a person in the loop, and what still requires one. Within the decision and operational layers of that architecture, most companies still have exactly one lever: human review, on or off, applied uniformly regardless of stakes or track record. A usable version isn’t a single switch. It’s three tiers.

Tier 1, human-in-the-loop. The system recommends, a person decides before anything executes. High-consequence, low-frequency, hard-to-reverse decisions belong here, regulatory filings, large pricing exceptions, anything with real legal exposure attached.

Tier 2, human-on-the-loop. The system decides and acts, a person monitors and can intervene inside a defined window before the consequences compound. Most operational workflows land here once you’ve actually built some track record with the system.

Tier 3, human-out-of-the-loop, bounded. The system acts autonomously inside an explicit blast radius, a capped dollar amount, a reversible action, a pre-cleared category. A person audits the pattern, not the transaction.

Toyota never gave a worker unlimited authority. They could stop the line. They couldn’t redesign the factory, renegotiate with a supplier, or change the product roadmap. The authority had a blast radius, which is exactly what these tiers are designing. They’re a tool for two of the five layers I’ve written about before, decision permission and operational permission, not a replacement for the other three. A perfectly bounded Tier 3 can still fail if whoever audits it shares the same blind spot the system does, or if the outside world, regulators, customers, counterparties, never extends the market trust the internal architecture assumed it would have.

The actual design work isn’t picking a tier once and moving on. It’s the mechanism that promotes a decision from Tier 1 toward Tier 3 as the system earns trust in that specific domain, and demotes it the moment it doesn’t, the same kind of circuit breaker I’ve argued multi-agent systems need for cost, applied here to authority instead of spend. Almost nobody has built that mechanism.

Toyota’s cord and Fannie Mae’s rulebook both had an advantage this version doesn’t. A violation was visible. A worker could see a defect. An underwriting file either met the rule or it didn’t. A model that’s slowly drifting in quality doesn’t trip a wire, it just gets quietly worse, and the same review process built to catch it can end up sharing its blind spot, the exact failure I’ve written about in automated underwriting’s own verification layer. The promotion and demotion mechanism above only works if an organization can actually detect drift in a probabilistic system, and that detection problem is harder here than it ever was on a factory floor. Building a Tier 3 that looks bounded on paper without solving that problem first is how the blast radius stops meaning anything.

The reason this becomes a durable advantage, and not just a nice-to-have, is that permission architectures are hard to copy. A competitor can buy the same model. They can license the same software. They can hire the same consultants who hired the same consultants. What they can’t buy off the shelf is the accumulated trust, the operating data, and the specific decision boundaries that let one organization move with confidence while another is still in a meeting arguing about who has the authority to approve the meeting’s outcome.

What Autonomy Actually Costs When You Get It Wrong

It would be dishonest to make this argument without naming what it costs when it’s done badly. Expanding autonomy without a real governance mechanism doesn’t remove risk, it just changes its shape, from slow and visible to fast and compounding. A bad approval chain produces one bad decision at a time, and someone downstream usually catches it. A badly bounded Tier 3 produces the same bad decision at machine speed, thousands of times, before anyone notices anything’s wrong.

That’s not an argument against building this. It’s the argument for building it on purpose instead of drifting into it. The companies that get hurt in this transition won’t be the ones that moved too slowly on autonomy. They’ll be the ones that expanded it without a defined blast radius, without an audit trail, and without a name attached to who owns it when it fails. Tiering, bounding, and auditing is the whole difference between deliberate autonomy and an accident waiting for a headline.

Three Questions Worth Asking This Quarter

If you’re the one accountable for this inside your company, the diagnostic isn’t how many pilots you’re running. It’s:

Where does a decision still require a human sign-off purely out of habit, not because the stakes or the risk actually call for it?

For your highest-volume automated processes, do you have a defined blast radius and an audit mechanism, or just an on/off switch?

Who owns the outcome when a system acts on its own and gets it wrong, and do they know yet that it’s their job?

Toyota moved authority from headquarters to the factory floor. Mortgage underwriting moved authority from individual judgment to an institutional system. AI is going to move authority from human execution to bounded autonomous systems. The technology changes each time. The pattern underneath it doesn’t.

Most companies can’t answer the second or third question today. That gap, not model capability, will separate the companies that become the default this decade from the ones that spend it running increasingly sophisticated pilots.