The 'Fragility Trap': Why AI Alone Won't Build Resilient Businesses

Investing in artificial intelligence without fundamentally rethinking organizational structures and decision-making processes can lead to improved individual metrics but ultimately leave companies more fragile in the face of disruption.

The Paradox of AI Investment and Resilience

Companies globally are pouring substantial resources into artificial intelligence, driven by the promise of enhanced efficiency, predictive capabilities, and deeper insights into market dynamics. Indeed, many firms report significant successes, leveraging AI for more accurate demand forecasting, refined customer understanding, and increased team productivity. However, a recent study, co-authored by Nina Shariati, founder of Circular Transparency, and Rickard Sandberg from the Stockholm School of Economics’ Center for Data Analytics, highlights a critical disconnect: these advancements often fail to translate into greater organizational resilience. The research, based on interviews with senior leaders from major global entities including H&M Group, ICA Gruppen, Ingka Group, Microsoft, Google, AWS, and BCG, reveals a concerning pattern where isolated improvements do not strengthen the enterprise as a whole, potentially making it more vulnerable to external shocks.

The core of the issue lies in how AI is integrated into existing business frameworks. While individual departments or processes might become acutely optimized, the overarching systems that connect these parts often remain unchanged. For example, AI can dramatically improve a retailer's ability to predict consumer demand with high precision, enabling faster commercial decisions. Yet, this newfound foresight frequently hits a bottleneck. Upstream functions such as sourcing, manufacturing, logistics, and supplier relationship management often continue to operate on legacy systems and slow, disconnected processes. The result is a company that excels at knowing what it needs but struggles to deliver it efficiently. As executives articulated during the study, seeing further into the future offers little advantage if the rest of the business cannot accelerate its response to meet that enhanced visibility.

The Unintended Consequences of Isolated Optimization

Numerous leaders interviewed for the study described a consistent scenario: AI tools sharpened their forecasts and customer insights, but the benefits did not propagate throughout the supply chain. This failure occurred because the supporting infrastructure—suppliers, planning mechanisms, and incentive structures—were not simultaneously redesigned to keep pace with the technological advancements. A common outcome is that one team or function becomes significantly smarter and more efficient, while the adjacent teams or processes remain static. The critical handoff points between these disparate parts then become areas where potential value is lost, creating internal friction rather than synergistic improvement.

This phenomenon explains why many organizations experience a peculiar paradox. While individual performance metrics improve, forecasts gain accuracy, and teams operate with greater efficiency, the business entity as a whole does not feel substantially stronger or more adaptable. The perception of enhanced capability at a micro-level does not translate into systemic resilience. The underlying problem is not with AI itself, but with the approach to its implementation. Many companies are treating AI as another IT project, something to be installed within existing structures, rather than as a catalyst for fundamental organizational transformation. The latter, however, demands a more profound shift, involving people, revised incentive structures, and a re-evaluation of how authority and decision-making responsibilities are distributed across various teams. This human and organizational element is far more challenging to address than merely deploying new technology.

The Ripple Effect and Emerging Vulnerabilities

As AI becomes more deeply embedded in business operations, the decisions made within one team increasingly create ripple effects across other areas. Effective decision-making now requires a holistic understanding of how diverse elements—customers, suppliers, technology, regulatory landscapes, and internal incentives—interact. It is no longer sufficient to focus solely on the operations of individual functions or the technical intricacies of the AI itself. This need for interconnected understanding arises against a backdrop of increasing global complexity. Supply chains remain vulnerable to geopolitical tensions, trade disputes, and unforeseen disruptions. Regulatory requirements are continuously expanding, cybersecurity threats are escalating, and climate-related events are becoming more frequent and severe. Concurrently, customer expectations continue to rise at an unprecedented pace, demanding ever-faster and more personalized responses.

The research also brought to light new vulnerabilities emerging from the very nature of AI adoption. One executive highlighted the increasing reliance of retailers on a limited number of dominant American cloud providers and foundational AI models. What once seemed a straightforward technological choice is now evolving into a strategic concern about control and potential exposure to single points of failure. Furthermore, the resource intensity of AI itself poses a growing challenge. Data centers, the indispensable infrastructure powering AI, consume vast amounts of energy and water. Several European regions are already experiencing significant strain on these resources, indicating that environmental and sustainability concerns, traditionally associated with distant manufacturing hubs, are now surfacing much closer to home.

The 'Fragility Trap' and the Path to True Resilience

Synthesizing these observations, a clear pattern emerges: companies are becoming adept at optimizing discrete parts of their operations, yet this progress often comes with a hidden cost. Such optimization can inadvertently lead to reduced flexibility, diminished spare capacity, and less room to adapt when circumstances change. While conditions remain stable, performance metrics may appear robust. However, when disruptions occur, performance can plummet dramatically. The study's authors term this dynamic the “fragility trap” – a state where efficiency gains at the micro-level inadvertently erode macro-level resilience.

Critically, AI is not the root cause of this fragility. The issue stems from the deployment of powerful new tools into organizational structures that were designed for a more predictable and slower-paced world. The businesses that will ultimately thrive in this evolving landscape will not necessarily be those that extract the maximum possible efficiency from AI in isolation. Instead, success will favor those willing to invest in the systemic flexibility required to pivot and adapt as conditions inevitably shift. This necessitates a fundamental re-evaluation of how decisions are made, clarity on accountability across teams, and a reconsideration of what metrics a company truly measures and rewards.

This perspective, originally presented in a report from Reuters, emphasizes that while AI offers immense potential for business improvement, its transformative power can only be fully realized when accompanied by deliberate organizational change. Without addressing the underlying structures, companies risk building systems that are highly optimized yet inherently brittle, leaving them unprepared for the volatility of the modern global economy.

Why it matters

For those in infrastructure, data centers, and telecommunications, this research underscores the critical need for systemic thinking. The increasing resource demands of AI, particularly in energy and water for data centers, represent a tangible resilience risk that can impact operational stability and sustainability goals. Furthermore, the reliance on a limited set of cloud providers highlights potential concentration risks within the digital infrastructure supply chain. Building truly resilient systems demands not only optimizing individual components, like high-performance computing or network throughput, but also ensuring the entire ecosystem, from power grids to fiber routes and cloud architecture, can adapt to unforeseen changes and support integrated business transformation. Ignoring these broader implications risks creating a highly efficient but ultimately vulnerable infrastructure for the AI-driven economy.