Redesigning Management Foundations that Integrate Growth and Trust Rebuilding the Foundations of Competitiveness and Trust for the Intelligent Systems Era

AI agents are the most likely to see broad adoption by 2030, with the potential to significantly raise the productivity of knowledge work(1). They can accelerate decision-making and execution, operate continuously beyond human limits, and create value not only in offensive domains but also in defensive functions such as security operations, audit, and anomaly detection. Physical AI is likely to spread more gradually across manufacturing, logistics, maintenance, construction, and healthcare or caregiving support, with the potential to drive a broader reconfiguration of physical labor around 2030(2).

The timeline for quantum computing requires careful interpretation. While generative AI has shifted some private investment, funding from governments, public funds, and leading corporations remains strong and is often accelerating. Competition among major countries and regions is intensifying. Quantum computing is increasingly viewed as a “strategic sovereign technology,” underpinning future industrial competitiveness and national security, and could, by around 2035, become a new source of differentiation in areas such as drug discovery, materials science, optimization, and complex calculations in finance and logistics(3).

What matters, however, is that the same capabilities that create competitiveness also create new sources of risk. These risks are not simply extensions of conventional cyber threats; they differ in speed, scale, attack surface, and the forms of damage they can produce.

AI agents are not only the greatest new level of productivity; they also introduce a new class of internal control risk. Runaway permissions, automated propagation of flawed decisions, data leakage, prompt injection, and cascading failures across agents are all emerging concerns. At the same time, external threats are evolving in kind. As AI capabilities diffuse, attackers gain similar advantages, accelerating the automation, sophistication, and scale of attacks. Language-agnostic, always-on campaigns, multimodal disinformation, and supply-chain-based intrusions are exposing enterprises to broader and more persistent risks than ever before.

However, the critical point is that the very capabilities driving competitiveness are also becoming new sources of risk. These risks are not a linear extension of traditional cyber threats; they differ in speed, scale, targets, and impact.

AI agents, through privilege escalation, automated propagation of errors, data leakage, and cascading failures across agents, can act as both a powerful productivity lever and a new form of internal control risk. Moreover, AI agents often hold credentials such as API keys, tokens, and SSO access, and can interact with multiple SaaS and internal systems. If compromised, they can become entry points for privilege abuse and lateral movement. Beyond isolated misuse, interconnected agents can serve as attack pivots, enabling risks to propagate and amplify across systems and environments.

Equally important is the qualitative shift in external threats. The widespread adoption of AI is equipping attackers with similar capabilities, accelerating the automation, sophistication, and scale of attacks. Threats are no longer constrained by language or time, and include data poisoning, prompt injection, multimodal disinformation, and supply chain infiltration—exposing enterprises to broader and more persistent risk landscapes.

What this implies is a fundamental shift: risks should no longer be viewed as isolated events, but as continuous processes—from attacker preparation and initial compromise to post-breach propagation and impact.

Physical AI presents an even more acute risk. Through robot takeover, sensor spoofing, breakdowns at the IT/OT boundary, disablement of safety mechanisms, and supply chain compromise, cyber risks can translate directly into physical damage and human safety incidents. Unlike traditional cyber risks—such as data breaches or system outages—malicious manipulation in physical environments can result in real-world accidents and injuries, representing a fundamentally different order of impact.

While quantum computing holds significant promise, it also raises substantial concerns(4). The accelerating pace of development is intensifying fears of a “Q-Day,” when widely used encryption could be broken, posing a near-term threat to digital security. The transition to post-quantum cryptography (PQC) and the inventory and remediation of cryptographic assets will take considerable time. Rather than waiting for large-scale business applications to materialize, organizations must act now from a quantum security perspective, anticipating the obsolescence of existing encryption. In particular, the “harvest now, decrypt later” threat—where attackers collect encrypted data today for future decryption—has already become a growing risk, especially for organizations holding long-lived sensitive data. Despite this urgency, most organizations remain underprepared for the transition to a post-quantum world.

In short, these three technologies are not only new drivers of competitiveness; they are also reshaping the foundations of trust, security, and resilience on which future competitiveness will depend.

Second, the redesign of work and operating models.

AI agents are entering knowledge work, while physical AI is transforming frontline activities, shifting enterprises toward execution models that integrate not only humans, but also digital and physical AI workers (see Figure 2). Implementation will begin in areas such as design, procurement, quality, maintenance, and back-office functions, ultimately requiring a reconfiguration of end-to-end workflows that connect IT and OT, as well as cyber and physical domains.

At the same time, this integration introduces new risks. As the IT/OT boundary blurs, cyber risks can propagate into the physical domain, leading to robot malfunctions or the disabling of safety mechanisms, with direct implications for human safety. From an attacker’s perspective, realistic entry points include bypassing safety shutdowns, weak or unclear human access controls and accountability, and infiltration through maintenance vendors. These risks underscore the need for safety-by-design in operating models where physical AI and humans work together.