Part I: Local Optima
I recently came across an email from Microsoft CTO Kevin Scott explaining internal tensions at OpenAI. And what led to the ousting of Sam Altman. He wrote:
“I could tell you stories like this from every place I’ve ever worked, and it boils down to, even if you have two important, super successful things you’re trying to work on simultaneously, folks rarely think about the global optima. They believe that their thing is more important, and that to the extent that things are zero-sum, that the other thing is a cause of their woes.”
That dynamic is universal.
Large systems are built from smaller systems that work. But strong functions do not automatically translate into organization-wide performance.
I used to wonder why.
The culprit is the Myth of Functions.
It is counterintuitive. But when competent people optimize for their local goals, they constrain the business. Sales optimizes for top-line revenue. Finance optimizes cost control. Marketing optimizes lead volume. Operations optimizes for unit efficiency.
Each leader defends a pristine dashboard. Each function improves its own metrics.
Locally, performance rises. Globally, coherence falls.
What is optimal for the part is often parasitic for the whole.
Part II: Local Optima and Politics
Local optimization persists because it is politically rational.
Power is vertical. Career progression is vertical. Functions are not only skill containers, but are also power structures. Within that context, it is rational to win locally even if the system suffers.
I recently reviewed a business where the friction was visible from the parking lot.
Consider what was happening inside their order-to-cash cycle, the subsystem that carries a customer from signed contract to collected revenue.
Sales closed deals with non-standard payment terms to hit quarterly targets. Finance, protecting margins, flagged those contracts for review, adding three to five days before invoicing even began. Operations, measured on unit efficiency, batched fulfillment weekly instead of processing orders as they arrived. Logistics, optimizing route density, held shipments until trucks were full.
Each decision made sense locally. Together, they broke the promise. The customer who signed expecting delivery in two days received their order in ten. No one violated policy. No one failed their KPI. Every function did its job. Yet the system failed.
Then we looked at how their OKRs were structured. Sales targeted quarterly revenue with no constraint on deal structure. Finance targeted DSO reduction to 14 days, penalizing slow-collecting contracts. Operations targeted a 15% improvement in fulfillment cost per unit, rewarding batching. Each OKR was reasonable in isolation. Together, they created a system where the customer’s needs became an afterthought.
The incentives required internal competition. For one team to hit its targets cleanly, another team had to absorb friction. Star performers were emerging by exploiting bottlenecks, not by resolving them. A star performer in a broken system is often just a symptom of a deeper design failure.
Leadership had engineered it. Because what you measure is what you manufacture. What you reward is what you reproduce. If you want folks to optimize for the system, you must bake in incentives that reward that behavior.
Part III: Containers vs. Subsystems
The deeper issue is not that functions behave tribally. It is that we have misidentified what functions actually are.
We have confused functions with subsystems.
Functions are containers. Sales, marketing, supply chain, finance, HR, product, engineering, design. These are not systems. They are containers of specialized skill. Capability clusters. Professional tribes.
A subsystem is something different.
A subsystem is a sequence of interdependent processes that produces a defined outcome. It serves either a customer or another internal unit.
Procure-to-pay is a subsystem, from identifying a need to paying the supplier. Order-to-cash is a subsystem, from customer order to revenue collection. Lead-to-revenue. Issue-to-resolution. Concept-to-Launch. Customer Retention.
These subsystems cut horizontally across the organization’s functional silos. And they depend on coordinated handoffs. When you treat a function as a subsystem, you strengthen the department and weaken the delivery. You cannot optimize a container and expect a subsystem to improve.
Part IV: Designing for Flow
Businesses are organized around departments, but they win or lose on how value moves. Customers do not experience your structure; they experience a journey. Throughput is governed by the narrowest constraint in the value stream, not the weakest department.
Leaders often obsess over functional excellence and ignore subsystem integration. The result is a disconnected enterprise. The Myth of Functions is the illusion that increasing capacity automatically increases output.
It does not.
You can strengthen every function and still fail to improve global performance. Most inefficiencies live in the white space between functions, the handoffs that the org chart does not show.
Big systems are built from smaller systems that work. But those smaller systems are not departments. They are value streams. If you want systems thinking to work, you must design around those streams.
Map the Journey. Trace the full path from trigger to outcome, ignoring departmental lines. A lead-to-revenue stream, for example, begins not when Marketing qualifies a prospect but when the first signal of intent appears, and it ends not when Sales books the deal but when Finance collects the cash. Until you draw the full path, you cannot see where value stalls.
Assign Stream Accountability. Give someone the authority to fix the handoffs. Someone whose incentives are tied to end-to-end cycle time rather than any single function’s dashboard. Without this, handoff failures become orphan problems that every function acknowledges, and no function owns.
Measure the Flow. Prioritize cycle time, quality, and throughput over how busy each function appears. Utilization tells you how much of a team’s capacity is occupied. It says nothing about whether that activity advances the journey. A team running at 95 percent capacity can still be the bottleneck.
Align Incentives. Reward shared outcomes so that “winning locally” feels like losing. For example, the VP of Sales is no longer measured on booked revenue but on collected revenue. Now she shares accountability with Finance and Operations. The Head of Operations is not rewarded for unit efficiency alone but for delivery time and customer satisfaction. Now efficiency cannot come at the expense of experience.
When incentives shift from departmental metrics to system outcomes, behavior follows. If compensation and prestige are tied to flow health, collaboration becomes rational. If they remain tied to silo performance, tribalism will persist. You cannot preach alignment and reward fragmentation.
Design influences behavior.
When you do this, functions assume their proper role. They become service providers to subsystems. Functions contribute expertise. Subsystems create outcomes. In the end, the market does not reward siloed excellence. It rewards synchronized execution.
The tragedy of the “Myth of Functions” is that rational behavior at the desk level often creates irrational outcomes at the enterprise level.