The AI Acceleration Effect. When Decision Velocity Exceeds Governance Capacity.

There are moments in markets when something subtle changes first, and only later becomes obvious to everyone.

I remember sitting on a trading floor in London in the late 1990s, watching execution speeds improve almost overnight. What had previously required layers of confirmation, discussion, and hesitation suddenly became immediate, orders moved faster, information travelled faster, decisions followed accordingly.

At first, it felt like pure progress. But, It wasn’t. What changed was not just speed. What changed was the distance between analysis and action. As that distance compressed, something else began to erode, quietly, the conditions under which judgment is formed. Decisions were still being made, but the space in which they could be challenged, reframed, or resisted was disappearing.

We are seeing the same dynamic again today, only this time, at a far greater scale.

Artificial intelligence is accelerating capability across industries at a pace rarely seen in previous technological shifts. Analysis that once required weeks now takes minutes. Strategic options can be generated, tested, and implemented almost instantly. Code is written automatically. Capital is deployed with increasing algorithmic precision. Entire operating systems are compressed into cycles that approach real time. The prevailing narrative is straightforward, more intelligence, more efficiency, more productivity.

However, that narrative is incomplete. Artificial intelligence, in spite of increasing efficiency, can also increase risk, but not in the way most assume. It increases the speed at which hidden fragility is exposed, and when decision velocity exceeds governance capacity, that exposure does not occur gradually, it can occur abruptly.

Every major technological shift has compressed time. The telegraph reduced communication delays, the internet reduced distribution friction, cloud computing reduced infrastructure constraints and artificial intelligence compresses something more fundamental, cognition.

The gap between analysis and execution is collapsing, what once unfolded across days or weeks now unfolds in minutes, this creates a powerful illusion, that faster decisions are inherently better decisions.

They are not.

Speed removes friction, and friction, in complex systems, has never been a defect, it has been a stabiliser. It slows capital allocation, allows dissent to surface, and creates the temporal distance required for second-order consequences to emerge. When friction disappears, systems do not become safer, they become more efficient at accumulating risk without recognising it. This is not a new phenomenon, it is structural.

In complex systems, risk does not accumulate linearly. It builds across layers, operational, financial, behavioural, and strategic, often remaining latent until a trigger reveals what was already embedded. What appears as a sudden disruption is usually the visible expression of a much longer process of silent accumulation.

The patterns differ in scale and narrative. The underlying structure does not.

Risk propagates through systems in a way that is recursive, interconnected, and highly sensitive to compression. When the time between decisions collapses, the ability to detect structural weakness diminishes with it and fragility does not simply increase, it becomes harder to see until it is already systemic.

I saw a version of this again years later, during the crisis of the early xxi century, not in markets but in boardrooms. Companies developed highly sophisticated, data-driven operating models. Monitoring could be adjusted dynamically, trading and risk management strategies could be continuously optimised, and operational changes could be implemented in near real time. On paper, it represented everything modern organisations aspire to achieve, in practice, something else was unfolding.

Decisions were being made continuously, but not always examined with the same depth. The cadence of execution had accelerated dramatically, but the cadence of governance had not.

At first, this went unnoticed. The performance was strong. Momentum was positive. Confidence was rising.

But beneath that momentum, assumptions were no longer being challenged with the same rigor. Strategic patience shortened. Dependencies increased. Organisations were not becoming reckless, they were becoming compressed. That compression is where fragility forms. Not through visible excess, but through the accumulation of decisions that are individually rational and collectively unstable.

Modern organisations now operate with the capacity to act continuously. Capital can be deployed instantly, products launched daily, and strategy adjusted in real time. The operational advantage is obvious. The structural implication is less visible.

As decision velocity increases, error does not increase linearly. It propagates through interdependence.

Each decision influences multiple layers simultaneously, financial exposure, operational dependency, market positioning, and behavioural response. When decisions are made faster than they can be fully understood, the system begins to reinforce its own assumptions. Feedback loops tighten. Correction mechanisms weaken. What emerges is not simply faster execution. It is a system that becomes increasingly confident in assumptions and structures it has not had time to validate. This creates a fundamental mismatch, a governance-velocity gap. Execution has moved to real time. Oversight remains periodic and slow.

Boards were designed for cadence, quarterly reviews, structured reporting, committee cycles. These frameworks assumed a world where change unfolded more slowly and decision points were discrete.

AI-enabled systems do not operate that way. Iteration is continuous, feedback is immediate, competitive responses occur in hours, not months.

A board that meets four times per year cannot meaningfully supervise systems that evolve daily without redesigning its architecture. The consequence is not immediate failure. It is the accumulation of silent fragility across layers, until a trigger forces the system to adjust faster than it can adapt.

In previous cycles, fragility accumulated primarily through financial leverage. In this cycle, it accumulates through decision compression.

When organisations can act continuously, they can also overextend continuously, without the traditional signals that would have triggered caution. Instability is no longer concentrated, it is distributed, embedded, and temporarily masked by momentum.

Capital markets are already reflecting this dynamic. Artificial intelligence requires extraordinary infrastructure investment, computational capacity, data centres, energy consumption. These introduce new forms of concentration, both financial and structural.

At the same time, capital is pursuing AI exposure with extraordinary intensity, valuations expand, funding rounds escalate, time horizons compress, the narrative feels new, the structure is not.

We have seen variations of this before, periods where capital flows into infrastructure ahead of discipline, where underwriting standards deteriorate under momentum, where the velocity of capital formation exceeds the capacity to govern it.

The surface changes. The geometry repeats. Confidence expands. Capital accelerates. Governance lags. Adjustment follows.

None of this is an argument against artificial intelligence, the opportunity is real and the transformation is profound. However, acceleration without structural reinforcement does not produce resilience, it produces faster exposure to underlying weakness. Artificial intelligence systems optimise for defined objectives under assumed parameters. They do not inherently question whether the objective is misaligned, whether the parameters are incomplete, or whether the environment has shifted. That function belongs to judgment. Moreover, judgment is not simply analytical, it is contextual, it integrates history, recognises patterns, tolerates dissent, and resists narrative momentum, and critically, it does not scale with processing power.

For decades, leadership advantage was partially based on access to information. Artificial intelligence is reducing that asymmetry, analysis is becoming commoditised and strategic simulation is becoming widely available.

As intelligence becomes abundant, discernment becomes scarce. The differentiator will not be who has more information, it will be who understands how decisions propagate through systems, and who has the discipline to slow down when acceleration becomes dangerous.

Acceleration does not create strength. It magnifies structure.

If the underlying system is robust, acceleration compounds advantage. If it is fragile, acceleration compounds failure.

The response is not to slow down blindly, it is to redesign governance.

Boards must increase engagement frequency where AI exposure is material. Risk frameworks must evolve from static reporting to dynamic system awareness. Capital allocation must be stress-tested not only for growth, but for constraint, regulatory intervention, energy limits, and shifts in narrative.

Founders must distinguish between speed and durability. The ability to move fast does not eliminate the need for disciplined balance sheets, aligned incentives, and strategic patience.

The critical question is no longer how fast an organisation can move, it is whether its governance architecture can sustain the speed at which it moves, because in an accelerated system, failure rarely comes from lack of intelligence, it comes from acting on that intelligence before it is understood.

In previous cycles, failure often came from not seeing risk, in this one, it may come from seeing everything, and acting before we understand what we are seeing.

Resilience, in that environment, is not a function of capability, it is the consequence of design, and design, as always, has to be deliberate.