Matilda Bailey is a distinguished networking specialist whose work sits at the intersection of cellular innovation and next-generation industrial solutions. With years of experience navigating the complexities of high-stakes environments like manufacturing and utilities, she has become a leading voice on how to integrate artificial intelligence without compromising the integrity of critical infrastructure. In this conversation, we explore the dual nature of AI as both a security risk and a defensive powerhouse, the cultural hurdles of IT/OT convergence, and the infrastructure shifts necessary to support a future of autonomous machine-to-machine decision-making.
Security is currently the top obstacle to AI adoption in industrial sectors, yet there is a strong belief it will eventually improve overall safety. How do you reconcile these conflicting views, and what specific steps can teams take to move from fear to functional defense?
It is a fascinating paradox because while 40% of professionals cite security as a major barrier, a massive 85% actually expect AI to eventually strengthen their defensive posture. We reconcile this by recognizing that AI is a double-edged sword; it introduces new vulnerabilities during the implementation phase, but it offers the only way to achieve the speed and scale required for modern threat detection. To move from fear to functional defense, organizations must treat cybersecurity as a baseline requirement rather than a downstream control. This involves moving toward secure-by-design architectures where 48% of teams are already identifying security as their primary networking challenge. By investing in AI specifically to improve cyber resilience, teams can transform the network from a point of vulnerability into an adaptive, self-healing system.
Only one-fifth of industrial organizations report full collaboration between IT and OT teams on cybersecurity. What are the primary cultural friction points causing this divide, and how does this lack of alignment specifically impact the speed and repeatability of scaling AI projects?
The friction usually stems from a fundamental difference in priorities, where IT focuses on data integrity and OT prioritizes uptime and physical safety, leading 43% of organizations to operate with limited or no cooperation. When these teams remain siloed, the visibility of cyber risks is diminished, which is a critical problem because scaling AI is as much an organizational challenge as it is a technical one. Without alignment, companies struggle with the repeatability of their projects, as there is no unified framework to move a pilot program from a single factory floor to a global enterprise. In contrast, full collaboration allows for a “force multiplier” effect where security gaps are identified early, ensuring that AI deployments are both stable and secure from the outset.
Integrating AI technology has overtaken labor shortages as a primary challenge for industrial networks. Why has the focus shifted toward integration, and what specific technical hurdles must be cleared to ensure new AI tools interact safely with sensitive legacy infrastructure?
The shift toward integration indicates that the initial “talent gap” panic has stabilized, and companies are now getting their hands dirty with the actual technical deployment. We are seeing that 61% of respondents are actively deploying AI, but only 20% have reached a mature, scaled adoption because the technical hurdles are immense. The primary challenge is ensuring that high-bandwidth AI workloads do not disrupt the deterministic nature of legacy industrial systems that were never designed for such data density. To clear these hurdles, organizations must modernize their connectivity to handle the increased telemetry without causing latency issues that could lead to physical equipment failure. This requires a transition to more robust, AI-ready environments that can bridge the gap between decades-old hardware and cutting-edge software.
High-speed wireless reliability and edge compute capacity are now seen as foundational requirements for industrial AI. What specific hardware upgrades are necessary to handle these heavy workloads, and how should organizations balance performance needs against the power constraints of traditional industrial network designs?
The demand for hardware has shifted dramatically, with 96% of professionals stating that wireless reliability is now the absolute foundation for enabling industrial AI at scale. To support these workloads, we are seeing a push for greater edge compute capacity, cited by 44% of teams, and increased bandwidth, cited by 42%. Organizations must upgrade to high-performance edge gateways and localized servers that can process data on-site to reduce latency, but they must do so while managing the significant power draw these units require. Balancing this involves redesigning industrial networks to accommodate higher energy consumption and heat dissipation, as 97% of respondents expect these new workloads to fundamentally change their traditional network design assumptions.
Advancing AI requires a shift from human-in-the-loop workflows to autonomous machine-to-machine decision-making. What operational risks arise when removing human oversight, and what data infrastructure improvements are needed to ensure these autonomous systems remain resilient and predictable?
When you move to machine-to-machine decisioning, the primary risk is the loss of a “sanity check” on anomalous commands, which could lead to cascading operational failures if the system is fed corrupted or malicious data. To mitigate this, we need massive improvements in data infrastructure to ensure that the information being shared between machines is accurate, timely, and securely encrypted. Investment in mobility and high-speed connectivity is essential here, as 51% of respondents are already anticipating significant increases in these requirements. Building resilience into these autonomous systems means creating a network that can monitor its own health and revert to safe states automatically if the AI identifies a deviation from predicted patterns.
Threat actors are using generative AI to automate phishing and malicious code development. How can defenders use AI-driven analytics to process vast amounts of network data, and what steps should they take to shorten the time between detection and response?
The reality is that threat actors are using generative AI as a force multiplier to increase the realism of social engineering and the speed of malware creation, making traditional defenses obsolete. Defenders must counter this by using AI-driven analytics to sift through massive volumes of network telemetry that would be impossible for a human to analyze in real-time. By identifying anomalous behavior—such as a sensor communicating with an unauthorized external IP—defenders can use automated response protocols to isolate the affected segment instantly. Shortening the detection-to-response timeline requires a move toward these automated, AI-augmented workflows that can act in milliseconds rather than hours.
What is your forecast for industrial AI?
I predict that over the next few years, we will see a rapid consolidation where the 20% of companies currently at mature adoption will begin to dominate their respective markets due to massive efficiency gains. As the focus shifts from pilot projects to full-scale machine-to-machine ecosystems, the “double-edged sword” of security will become less of a barrier and more of a competitive advantage. Those who invest early in collaborative IT/OT structures and robust edge compute hardware will create a resilient foundation that can withstand both labor shortages and increasingly sophisticated AI-driven cyberattacks. Ultimately, AI will not just be an add-on; it will become the very fabric of the industrial network, enabling a level of autonomy and safety that was previously unthinkable.
