Organizational Structures for P Dominance: What Does a Company Look Like When Judgment Is the Constraint?

In the last article, I wrote about hiring and structuring teams to handle P work.

But there’s a level above the team. The org itself.

And most organizations are built on a foundational assumption that’s quietly becoming wrong.

The assumption: the bottleneck is execution capacity.

That assumption was correct for most of industrial history. You ran short on workers who could execute. So you added people, added managers to coordinate them, added layers to coordinate the managers. Org charts grew taller. Reporting lines multiplied. Planning cycles extended. The whole architecture was optimized for one thing: moving D work through the system efficiently.

Now picture what happens when AI absorbs that D work.

You still have the hierarchy. The planning cycles. The reporting layers. The handoff structure. But the bottleneck has moved. The constraint isn’t execution anymore. It’s judgment — who’s making good calls under uncertainty, how fast, and at what quality.

And the org chart — designed for a problem that no longer exists — is actively in the way.

1) The D-Era Org Was Engineered to Move Work, Not Improve Judgment

It’s worth understanding why org structures look the way they do before trying to change them.

The classic hierarchy has a logic: - Senior leaders own strategy (the highest-stakes P decisions) - Middle management translates strategy into plans (lighter P + structured D) - Individual contributors execute the plans (mostly D) - Information flows up through status reports and metrics - Decisions flow down through directives and roadmaps

That’s a D-work delivery machine. It’s reasonably efficient at turning known goals into shipped outputs. The structure assumes: - what to build is figured out at the top - how to build it is figured out by management - whether it ships is determined by ICs executing cleanly

When D is the bottleneck, this works. More execution capacity at the bottom, better coordination in the middle, clearer goals at the top.

But watch what happens when AI automates execution.

The ICs aren’t executing D work anymore — AI is. They’re making judgment calls about what AI should do, evaluating what it produced, and deciding what to do with ambiguous results. Their work shifted from D to P without their job title, compensation, or reporting structure changing at all.

Middle management used to be responsible for translating strategy into work that could be executed. Now there’s less execution to coordinate. But there are more judgment calls at every layer — and the management structure isn’t designed to route, support, or improve those judgments.

And the senior leaders at the top? They used to be information-constrained: by the time data surfaced from the bottom, it was stale. AI has fixed that — they can see nearly real-time what’s happening. But if they’re still making the top-level P decisions while the organization underneath them is full of unresolved P work that nobody is accountable for, you have a judgment bottleneck at every layer, with no architectural solution for it.

The org didn’t break. The bottleneck moved. And the org hasn’t caught up.

2) The Two Structural Failure Modes

Before getting to what good looks like, let’s name what bad looks like. P-dominant orgs break in one of two predictable ways.

Failure Mode A: Centralized Judgment — The Bottleneck Becomes the Leader

Faced with a world full of uncertainty, many organizations default to centralizing decisions upward. Senior leaders review more. Approval chains lengthen. Nothing ships without sign-off from the person with the most context.

This feels safe. The highest-quality judgment is being applied.

What actually happens: - Decision latency increases. Fast-moving environments punish this. - Leaders become chronically overloaded with P decisions they shouldn’t be making - The team below loses judgment capacity — because they’re never asked to exercise it - Junior employees optimize for “what will get approved” instead of “what’s actually right”

You end up with a judgment monopoly. One or two people hold the cognitive model for how the company navigates uncertainty. When they’re right, great. When they’re wrong — or tired, or absent — there’s no backup.

The failure mode for P work is identical to the failure mode for any single point of failure in a distributed system: the bottleneck node collapses under load.

Failure Mode B: Distributed Autonomy — The Org Fragments Into Independent P Universes

The opposite reaction is to push decisions down everywhere. Teams get full autonomy. Move fast, trust your team, ship and learn. Decentralized judgment.

This also feels right. Local context is best for local decisions. Less bureaucracy, more ownership.

What actually happens: - Each team develops a private P model — their assumptions, their risk tolerance, their bet framing - These models diverge quietly over time - Strategic coherence erodes as each team optimizes locally - Nobody knows what the company believes about its own market, its customers, its bets - The company is simultaneously a dozen overconfident probabilistic experiments with no shared framework

You get the opposite of the bottleneck problem: not one P decision point, but dozens — uncorrelated, uncoordinated, and collectively incoherent.

The sum of individually rational P bets can be organizationally irrational. Two teams optimize for opposite user behaviors. Three teams build conflicting infrastructure assumptions into their product. One team’s kill criteria would contradict another team’s green-light criteria if anyone compared them.

3) The Right Question Isn’t “Flat vs. Hierarchical”

The standard org design debate in tech is about centralization vs. decentralization. Hierarchies vs. flat teams. Top-down vs. bottom-up.

That debate is about where decisions get made.

For P-dominant orgs, the more important question is: how does judgment get routed, improved, and shared across the organization?

Because the real problem isn’t whether decisions happen high or low — it’s that every decision is happening in isolation, with no mechanism for the organization to get smarter over time.

Think about what “judgment routing” would mean: - The right P decision reaches the person with the best combination of context + expertise for that specific uncertainty - The reasoning behind that decision is captured and made available to the rest of the org - When a decision turns out wrong, the organization can trace back to its assumption and update - When the same class of decision recurs, the org has accumulated wisdom about it instead of starting from scratch

Compare that to what most orgs do: decisions are made, things are shipped, postmortems happen occasionally, but nobody knows what was assumed, why, or whether those assumptions were validated.

Judgment doesn’t improve in most orgs because it’s not treated as an organizational asset. It’s treated as a personal attribute — someone is good at decisions or they’re not.

4) Structural Shifts for P-Dominant Organizations

Here’s what actually needs to change. These aren’t speculative — they’re patterns that are already emerging in companies that have figured out how to scale in uncertain environments.

A) Replace the Planning Cycle with the Assumption Cycle

Traditional org calendars are built around delivery planning: quarterly roadmaps, annual goals, sprints.

That’s a D-work calendar. You’re planning what to execute.

P-dominant orgs need a parallel cadence: the assumption cycle.

The assumption cycle answers different questions than the planning cycle: - What do we believe right now that’s carrying the most risk? - Which of our active bets have assumptions that have been tested? - Which haven’t been tested and should be? - What changed in context this quarter that invalidates something we were confident about?

This doesn’t replace planning. It runs alongside it. The planning cycle manages D-work execution. The assumption cycle manages P-work quality.

Most orgs have the first. Almost none have the second.

B) Judgment Infrastructure as a First-Class Org Function

In a D-heavy org, operations exists to make execution smoother: process, tooling, coordination.

In a P-heavy org, you need an equivalent function for judgment: making sure good P decisions are made, captured, shared, and learned from.

This might not need to be a dedicated team. But it needs to be someone’s explicit job. Judgment infrastructure includes:

• Decision log ownership — Someone maintains the record of what was decided, why, and what the assumed outcome was
• Assumption registry — A living document of what the organization currently believes, with confidence ratings and last-tested dates
• P work QA — Periodic review of decision quality: not just “did it work?” but “was the reasoning sound?”
• Checkpoint design — Ensuring high-stakes P decisions have defined review points before action

If no one owns this, it will not happen. It’s too easy for execution velocity to crowd out the slower, more uncomfortable work of examining your assumptions.

C) The New Communication Layer: Assumption Broadcasting

Most company communication is status reporting: what shipped, what’s blocked, what’s coming.

Status reporting is D-work communication. It tracks outputs.

P-work communication needs an additional layer: assumption broadcasting.

When a team makes a major bet, they should be broadcasting: - “We’re making this decision. We believe X.” - “Our kill criteria is Y.” - “We expect to know if this worked by Z.”

Not for approval. For coordination. So that other teams who hold adjacent bets can update their own assumptions if needed. So that if someone else’s context invalidates your assumption, they can flag it before you’ve built for three more months.

In a D-heavy org, teams communicate through handoffs. “I finished this, now you build on top of it.”

In a P-heavy org, teams communicate through belief updates. “Here’s what I believe right now, and here’s the evidence that’s changing it.”

That’s a fundamentally different communication protocol. And most companies have no channel for it.

D) Leadership Role Shift: From Decision-Maker to Judgment Amplifier

In the D-era, senior leaders added value by making better strategic decisions than the people below them. Their job was to be the highest-quality P node in the hierarchy.

When P work is distributed throughout the organization, that model breaks. Leaders cannot maintain the full context needed to make good decisions for every team. The organization is too fast, too complex, too context-dependent.

The new leadership role is to amplify judgment instead of monopolizing it.

What judgment amplification looks like: - Teaching framing skills, not just approving decisions - Creating the conditions where bad news surfaces fast (psychological safety as infrastructure, not as culture talk) - Modeling good reasoning in public — showing how you weigh tradeoffs, how you update, how you kill your own ideas when evidence changes - Building the assumption-sharing and decision-logging systems the org needs - Running calibration: are we collectively over-confident? Under-confident? What are we systematically missing?

The shift from decision-maker to judgment amplifier is uncomfortable for most senior leaders, because it requires giving up the role of oracle. But it’s the only model that scales.

A leader who is the organization’s sole P bottleneck caps the organization’s judgment capacity at their own. A leader who amplifies judgment can compound it across everyone.

5) How to Scale P Without Scaling Headcount

One of the most uncomfortable truths about P work: it doesn’t scale the way D work scales.

D work scales predictably. Need to process twice as many support tickets? Hire twice as many agents (or deploy twice as many AI endpoints). Output is roughly proportional to capacity.

P work doesn’t work like that. You can hire twice as many smart people and still have the same quality of judgment decisions — because judgment quality isn’t a headcount problem. It’s a context, calibration, and learning-loop problem.

So how do you scale P? A few mechanisms that actually work:

Accumulate decision capital

Every decision you make and document is an asset. When the same class of decision recurs, the org doesn’t start from scratch — it has a precedent, an outcome, a lesson.

Most orgs throw this away. Decisions are made in Slack, documented nowhere, and lost when the people who made them leave. The org makes the same framing error six months later, with a new team that doesn’t know the history.

Decision capital is the compound interest of P work. It grows if you invest in it. It’s zero if you don’t.

Institutionalize calibration

Calibration is the practice of comparing your confidence levels to outcomes. If you said you were 80% confident and you were right 40% of the time, you’re systematically overconfident. If you were right 95% of the time, you’re underconfident.

Individual calibration is a personal skill. Organizational calibration is a structural practice: periodically comparing what the org believed with what actually happened, and adjusting the collective confidence accordingly.

Calibrated orgs make better bets not because they’re smarter, but because they’re more honest about uncertainty. They don’t bet 80% on things that are 50-50. They don’t miss high-confidence opportunities because they’re locked in analysis.

Build judgment diversity into decision pods

Identical backgrounds produce correlated biases. If your decision-making group all went to the same schools, worked at the same companies, and were selected by the same hiring process, you have a portfolio of P decisions that all share the same blind spots.

Judgment diversity — different domain expertise, different experience with risk, different pattern libraries — is how you reduce correlated error. It’s not about representation as an optic. It’s about intellectual robustness as a structural property.

When your framing pod (see the last article) includes someone who’s run small companies, someone who’s worked in constrained resource environments, someone who’s failed in public — you get a broader P model than a pod of homogeneous high-performers.

6) The Communication Stack for P-Dominant Orgs

Here’s what the communication layer looks like when it’s actually built for P work:

The top row — status — is what most orgs have. Everything below is what most orgs are missing.

7) What Doesn’t Change

To avoid overcorrecting: a lot of the standard org design wisdom still applies.

Clear ownership still matters. P work doesn’t make ownership obsolete — it makes it more important, because ambiguity is costlier. Every high-stakes P bet still needs an explicit owner.

Execution discipline still matters. Shipping nothing is not a P strategy. The point of improving judgment is to make better bets, not to avoid making bets.

Culture still matters. Psychological safety, trust, low-drama communication — these are prerequisites for truth-telling in any environment, D or P.

What changes is the additional infrastructure layered on top. The assumption cycles, the decision logs, the judgment amplification model for leaders, the communication stack that includes belief updates not just status updates.

Think of it as adding a P layer on top of your existing D infrastructure. You’re not tearing down the org chart. You’re building the system the org chart was never designed to have.

8) Audit: Is Your Org Built for P Dominance?

Answer these honestly:

1. Assumption cycle — Does your org have a regular cadence for auditing what you currently believe and testing it against new evidence? Or does “review” mean reviewing output, not assumptions?

2. Decision capture — When a major bet is made, is the reasoning documented — not just the outcome? Can you trace a decision to its assumptions six months later?

3. Communication protocol — Do teams share what they believe (and why) across the org? Or does communication primarily flow through status reports about delivery?

4. Leadership model — Are your senior leaders the primary decision-makers for high-stakes bets? Or are they building the infrastructure for distributed judgment?

5. Calibration practice — Does your org compare confidence levels to outcomes at any scale? Or do you only find out what went wrong when it goes obviously wrong?

6. Judgment diversity — Do your key decision pods have meaningfully different P frameworks, or are they homogeneous in background and risk profile?

If you answered “no” to four or more: your org is running D-era infrastructure in a P-dominant environment. The misfit is real, and it’s costing you more than you can see from inside it.

Closing

When execution is the bottleneck, org structure is about coordination.

When judgment is the bottleneck, org structure is about amplification.

The companies that navigate this well won’t necessarily have the flattest hierarchies or the most autonomous teams. They’ll have the best systems for making judgment a compoundable, distributable, learnable org capability — instead of treating it as something a few smart leaders happen to carry in their heads.

That’s a different design problem than anything business schools taught in the D era. It doesn’t have clean textbook answers yet.

But the organizations that figure it out first will have something that’s genuinely hard to replicate: not a product advantage, not a distribution advantage — a judgment advantage. And in a world where everyone’s building with the same AI tools, that’s what actually compounds.

Part of the D/P Framework series. Previous: Building Teams for P Work. Next: The Economics of P. Building in public at fullstackpm.tech. Follow along on X @fullstackpmtech.