When Influence Replaces Control
The main misconceptions shaping AI systems
and leadership decisions.
Across AI systems, the same pattern appears repeatedly:
We assume control where we mainly have influence.
We assume clarity where the system creates ambiguity.
We assume agency where constraints shape the available choices.
What happens when leaders act on a system they fundamentally misread?
The five tensions below are active fault lines between what AI actually is and what most discourse assumes it to be. Each one carries identifiable misconceptions, the overall consequence and a diagnostic question leaders can use to test their own assumptions.
1. Acceleration vs Constraint
We celebrate speed, but constraints set the limits.
Misconception: “AI scales like software.”
Reality: It scales like industry, with physical, economic, and energy constraints. Our full white paper elaborate on this.
Misconception: “Compute will always catch up.”
Reality: Only if energy, capital, and supply chains allow it.
Misconception: “Progress is purely technological.”
Reality: Progress also depends on infrastructure, manufacturing, permitting, financing, and geopolitical access.
Consequence:
Leaders believe they are accelerating. In practice, they may be accelerating into constraints they do not control.
Diagnostic question:
Do we know which physical or infrastructural constraint is most likely to limit our AI roadmap in the next 18 months, and who controls it?
=> Boards need to understand where scaling depends on external systems that may become unstable, constrained, or politically exposed.
2. Concentration vs Diffusion
AI usage spreads widely. Control does not.
Misconception: “Open-source democratizes AI.”
Reality: It democratizes model access more than compute access.
Misconception: “More players mean less concentration.”
Reality: Concentration often increases at the infrastructure, cloud, and capital layers.
Misconception: “AI is globally distributed.”
Reality: Usage is distributed. Strategic control remains concentrated.
Consequence:
Leaders may believe they operate inside an open competitive landscape while depending on a small number of underlying actors.
Diagnostic question: Can we distinguish which parts of our AI capability we control, which we access through providers, and which we depend on actors we cannot influence?
=> Leadership should map dependencies below the application layer: cloud concentration, compute access and hyperscaler influence.
3. Interdependence vs Fragmentation
Friction increases. Interdependence remains.
Misconception: “AI will split into separate geopolitical worlds.”
Reality: Dependencies across chips, materials, infrastructure, and capital remain deeply interconnected.
Misconception: “Globalization continues unchanged.”
Reality: Friction, export controls, and strategic rivalry increasingly affect cost, speed, and access.
Misconception: “Decoupling solves dependency.”
Reality: It often replaces one dependency with another.
Consequence:
Leaders may underestimate how exposed their systems remain to external geopolitical dynamics.
Diagnostic question: If a key geopolitical friction point escalated tomorrow, which of our AI dependencies would be disrupted, and do we have a mapped alternative?
=> Scenario thinking must extend beyond markets and competitors. It must include infrastructure access, supply-chain fragility, export restrictions, and geopolitical spillover effects.
4. Competition vs Stability
Competition drives capability. It can also destabilize the system.
Misconception: “AI competition is simply a technology race.”
Reality: It is also a race for infrastructure, energy, talent, and strategic positioning.
Misconception: “More competition automatically improves outcomes.”
Reality: It can increase duplication, deployment pressure, and interoperability problems.
Misconception: “Fragmentation creates safety.”
Reality: Fragmentation can weaken coordination and increase instability between systems.
Consequence:
Organizations optimize locally while systemic coordination weakens globally.
Diagnostic question: Are we making AI deployment decisions based on competitive pressure alone, or do we also account for systemic effects on interoperability, standards, and shared infrastructure?
=> Leaders should evaluate not only competitive advantage, but also systemic exposure:
dependency accumulation, coordination failures, and deployment pressure created by the wider ecosystem.
5. Delegation vs Responsibility
Capability expands faster than accountability structures evolve.
Misconception: “Human in the loop guarantees responsibility.”
Reality: Humans often validate outputs without fully interrogating them.
Misconception: “AI only automates tasks.”
Reality: It increasingly shapes interpretation, prioritization, and framing.
Misconception: “Responsibility can be distributed across the system.”
Reality: Tasks distribute easily. Accountability does not.
Consequence:
Responsibility gaps emerge precisely where systems become most operationally effective.
Diagnostic question: For each AI system we operate, can we name the person accountable for its outcomes (not only its outputs)?
=> Organizations need visible points where judgment remains explicit:
who challenges assumptions,
who authorizes deployment,
who can override the system,
and who ultimately owns the consequences.
The real risk
Across all five tensions, the same dynamic appears:
AI systems evolve according to constraints, incentives, dependencies, and strategic pressure, and not only according to leadership intent.
The danger is overestimating how much control still exists over the systems shaping decisions.
Misconceptions are not only intellectual errors.
They become operational risks when strategic decisions are built on them.
Executive Reflection
Leadership increasingly operates inside systems that cannot be fully directed from any single point. In such environments, the diagnostic questions above are not rhetorical but are the minimum threshold for informed judgment.
If leaders cannot clearly identify which parts of the system they control, which they only influence, and which operate entirely outside their reach, the decisions based on that system are not fully understood.
AI systems change the conditions under which leadership operates.
In constrained and interdependent systems, misunderstanding the environment can become a strategic risk in itself.
Igor Allinckx
AI & Governance
June 2026
Back to Insights
Part of an ongoing exploration of governance, AI, and human judgment.