If We Can Measure Everything, Do We Still Need Politics?

In a previous piece — What About Teams? — I explored an idea that has been circulating more recently, including variations of Jack Dorsey’s model:

What if companies move away from stable teams entirely and operate as fluid networks of individuals, dynamically forming around work?

The intuition behind that model is straightforward:

• Teams are a coordination artifact.
• Coordination exists because of human limitations.
• If systems can coordinate better than humans, teams become optional.

And if you extend that idea just one step further — if the system can also track real performance, for everyone, continuously and objectively — then something more radical emerges:

Maybe we don’t just remove teams. Maybe we remove the need for hierarchy, friction, and even politics.

It is a seductive thought. It is also almost certainly wrong.

The temptation: perfect measurement kills politics

There is a compelling mental model here.

If we can track real contribution for everyone, continuously and objectively, then hierarchy, friction, and politics should simply disappear.

In that world:

• Everyone operates like an IC.
• Contribution is visible in real time.
• Recognition becomes automatic.
• Coordination becomes data-driven.

No more narrative battles. No more “who said it louder in the room.” Just signal. Clean, continuous signal.

If you have ever spent a performance review cycle watching someone take credit for a project they joined in the last two weeks, this vision has a certain emotional appeal.

The first break: measurement is not reality

Even with perfect instrumentation, we do not capture everything that matters.

We can measure:

• outputs
• throughput
• system impact
• dependency graphs

But we struggle to measure:

• preventing bad decisions early
• shaping ambiguous problems
• influencing direction without authority
• building trust inside a team
• long-term architectural taste

These are not edge cases. They are a large part of senior-level contribution. The person who quietly killed a bad migration before it started does not show up in any dashboard. The metric says they shipped nothing that quarter.

The system optimizes for what it can see. Everything else becomes invisible labor.

The shift: from politics to metric gaming

When performance becomes measurable, behavior adapts. That is not a risk — it is a guarantee.

You do not eliminate politics. You change its surface area.

Instead of narrative control, you get metric control.

Patterns emerge quickly:

• Optimizing for visible work over meaningful work.
• Avoiding ambiguity — because ambiguous problems are hard to score.
• Shaping tasks to look better in the system.
• Redefining what counts as “impact” in ways that favor your output.

This is Goodhart’s Law in its purest form:

When a measure becomes a target, it ceases to be a good measure.

The old politics was about who tells the best story in a meeting. The new politics is about who shapes the metrics. Arguably, the second version is harder to detect and easier to deny.

The next move: what if AI detects the gaming?

Now extend the system.

What if it does not just measure — but also detects manipulation?

AI flags suspicious task fragmentation, avoidance of complex work, inflated contribution signals, weak downstream impact. Every loophole gets closed. A self-healing system.

At first glance, this sounds like the answer.

At second glance, it sounds like building a surveillance state for your own employees. But let’s set that aside for a moment.

The second break: you are modeling intent

Detecting gaming means you are no longer measuring behavior. You are inferring intent.

And that is fundamentally harder.

The same pattern can mean opposite things:

Pattern	Innocent reading	Cynical reading
Breaking work into small tasks	Clarity and focus	Inflation
Focused, narrow work	Efficiency	Avoidance
High output	Productivity	Shallow work

The system must answer: why did this happen?

Good luck with that. Humans have been trying to answer that question about each other for thousands of years, and we are still not great at it.

The infinite loop

The system never stabilizes.

This is the same loop as fraud detection, spam filters, and ranking algorithms. No equilibrium. Only evolution.

Anyone who has worked on content moderation or ad fraud will recognize this immediately. The system gets smarter. The gaming gets subtler. Repeat forever.

A new risk: overfitting the organization

As detection improves, the system gets better at recognizing known patterns. But it also starts penalizing unknown patterns — work that does not fit the model.

You risk optimizing for predictable productivity while suppressing breakthrough thinking.

The engineer who disappears for three weeks and comes back with a prototype that changes the product roadmap does not look productive on day fourteen. A system optimized for consistent output would flag them.

The human layer becomes more important

At scale, something flips.

The better the system gets, the more critical human judgment becomes.

Humans are needed to:

• Interpret intent.
• Protect unconventional thinkers.
• Recognize invisible work.
• Override the system when it is confidently wrong.

Because the system is necessarily incomplete. Every measurement framework has blind spots. The question is whether someone with enough context is watching those blind spots — or whether the dashboard has become the only source of truth.

So what actually changes?

Not this: hierarchy disappears and politics disappears.

But this:

Authority shifts from position to validated contribution.

That is a meaningful change. It is also a much smaller change than “politics is over.”

If we can measure contribution, can we price it?

Once contribution is visible, a natural next question follows: what if compensation tracks it directly?

Instead of fixed salaries and periodic adjustments, you get continuous compensation tied to measured contribution. A live market of company value.

This is where the idea gets genuinely interesting — and genuinely dangerous.

The illusion of a single value function

Dynamic compensation assumes we can define company value clearly.

But value is multi-dimensional:

• Revenue
• Growth
• Reliability
• Quality
• Brand
• Market positioning

Different roles affect different dimensions. The person who improves system reliability by 40% and the person who closes a deal worth $2M are both creating value. Comparing them requires the system to decide what matters most.

That decision is not measurement. It is encoded judgment. And whoever writes that function has more power than most managers ever did — they just do not show up on any org chart.

The attribution problem

Most value in a company is collaborative and non-linear.

• Multiple teams drive outcomes.
• Early decisions shape future success.
• Avoided failures leave no trace.

Attribution becomes probabilistic at best. And “fairness” becomes a model, not a fact.

Anyone who has tried to attribute a successful product launch to individual contributors knows this. Was it the PM who shaped the spec? The engineer who built the prototype? The designer who insisted on changing the flow three days before launch? The answer is yes — but the system needs a number.

The time horizon problem

Some work pays off now. Some pays off later.

• Bug fixes are immediate.
• Infrastructure work is delayed.
• Mentorship is indirect.

So the system must price future impact today.

No clean solution exists. And every approximation encodes a bias — typically toward short-term, visible output.

Incentives still win

Tie compensation to metrics and behavior adapts. Measurable work dominates. Ambiguity gets avoided. Attribution gets contested.

Even with AI in the loop, incentives shape behavior. They always have. The mechanism changes. The dynamic does not.

Market vs team

This loops back to What About Teams?

Should a company behave like a market or a team?

Markets optimize for individual contribution, dynamic ownership, and efficient allocation. Teams optimize for trust, shared goals, uncredited work, and resilience.

Push too far toward individual fairness and you weaken cooperation. Why would anyone spend a week onboarding a new teammate if the system only rewards their own output?

A more realistic model

1. High-fidelity signals

Better visibility into contribution. Less reliance on narrative. But treated as signals, not verdicts.

2. Persistent teams

Trust as a multiplier. Shared context as infrastructure. The things that do not show up in dashboards but make everything else work.

3. Human interpretation

Context, correction, and judgment. Someone who understands the difference between “shipped nothing” and “prevented a disaster.”

4. Bounded compensation

Dynamic, but not fully fluid. Tied to both individual contribution and team outcomes. Because pure individual optimization is just freelancing with extra steps.

Where politics goes

Politics does not disappear. It moves.

From meetings and narratives to metric design, model assumptions, definitions of value, and attribution rules.

Quieter. But more powerful.

The person who decides what the model measures now holds more organizational power than most VPs. They just do not have a title for it yet.

The take

Measurement does not remove politics. AI does not remove gaming. Dynamic pay does not create fairness.

They all move the problem into system design.

The question was never “can we measure everything?” It was always “who decides what the measurements mean?”

Closing thought

A team was never just coordination.

It was a response to cognitive limits, uncertainty, and the need for trust.

Even in a fully instrumented world, that does not go away. If anything, it becomes more valuable.

Because the better we measure work, the more we realize what we cannot measure.