This episode introduces the AI Dependency Register, a strategic governance tool designed to help boards and audit committees manage operational reliance on artificial intelligence. While many organisations track their AI tools, few understand the resilience risks that emerge when business workflows become inseparable from specific models or suppliers. This document explains that unmanaged dependency often develops through repeated local use, potentially leading to systemic failure if a service is disrupted or modified. The proposed register addresses this by mapping the entire supply chain, including fallback options, accountability, and technical ownership. Ultimately, the source argues that the goal of modern governance is not to avoid AI, but to achieve acceptable dependency by design through total visibility.

Full article:

Narrated by NotebookLM

This episode explores how the National Health Service (NHS) can enhance its safety protocols by adopting the "just culture" model utilised by the Royal Air Force (RAF). This framework distinguishes between unintentional human errors caused by systemic flaws and reckless misconduct, ensuring that accountability is handled with fairness and consistency.

This episode introduces the "control tower" concept as a vital framework for managing the rapid proliferation of AI agents within large organisations like the NHS. It argues that without a centralised operating model, institutions risk facing digital sprawl, redundant tools, and significant security vulnerabilities. To address this, the source proposes a seven-stage lifecycle—spanning from initial discovery to eventual retirement—to ensure every autonomous system remains accountable and safe. The core philosophy is to centralise governance standards while allowing local teams the freedom to innovate within established risk-based guardrails. Ultimately, the text highlights that true AI capability depends on an organisation's ability to monitor, evidence value, and maintain oversight of its entire automated workforce.

Full article: https://www.linkedin.com/pulse/agent-factory-needs-control-tower-chris-parsons-6ybxe/

Narrated by NotebookLM

This episode explores the gap between technical access to generative AI and the actual realisation of organisational value. It argues that licensing software is merely the first step in a complex journey that requires redesigning workflows and fostering behavioural changes. Successful integration depends on six critical conditions, including high-quality data foundations, targeted use cases, and strong managerial support. Furthermore, the source emphasizes the "jagged frontier" of AI capability, warning that the technology can boost productivity in some areas while degrading it in others. Ultimately, leadership must shift focus from usage metrics to measurable outcomes to ensure that time saved by AI is redirected toward meaningful, high-value activities.

Full article: https://www.linkedin.com/pulse/after-licence-what-actually-determines-whether-ai-delivers-parsons-uhehe/

Narrated by NotebookLM

This episode explores the critical shift from AI assistants to autonomous digital agents and the urgent need for structured organizational accountability. It argues that agentic AI should be treated like a human employee, requiring a clearly defined role, supervision, and boundaries rather than being viewed as a mere software tool. Using the NHS as a primary example, the author introduces the "five rights of agentic delegation" to ensure these systems operate with the correct authority, data, and escalation paths. Ultimately, the source emphasises that human leaders must remain responsible for the outcomes of AI actions, ensuring that technological autonomy does not result in a loss of safety or oversight.

Full article: https://www.linkedin.com/pulse/from-copilot-co-worker-who-accountable-when-ai-agent-acts-parsons-ojqfe/

Narrated by NotbookLM

This episode examines how the NHS can enhance its newly established Leadership and Management Framework by adopting the systematic development strategies used by the Royal Air Force. While the NHS has succeeded in defining leadership standards, the author argues that it lacks the intentional career-long pathway found in military training. The RAF treats leadership as a distinct profession, preparing individuals for future responsibilities through continuous education and experiential learning before they assume new roles. In contrast, healthcare professionals often receive leadership authority based on clinical expertise without sufficient prior preparation. By shifting from a reactive model to a proactive system of leadership formation, the NHS could create a more resilient and strategic workforce. Ultimately, the comparison highlights that effective leadership is an institutional capability that must be cultivated consistently across an entire career.

This episode explores the strategic shift from pursuing the largest possible artificial intelligence models toward adopting right-sized AI tailored for specific public-sector needs. It argues that while massive frontier models offer groundbreaking capabilities, they also introduce significant financial, environmental, and operational costs that may be unnecessary for many tasks. By utilising smaller, specialised models, organisations can improve sovereignty and resilience while making system assurance and data security more manageable. The author suggests a portfolio approach where model selection is based on the minimum capability required to perform a task safely and efficiently. Ultimately, the source emphasises that operational control and sustainability are just as vital to national AI strategy as raw computational power.

Full article: []

Narrated by NotebookLM

This episode explores the transition of artificial intelligence from a mere productivity tool to a vital form of critical national infrastructure. It argues that when society becomes operationally dependent on these systems for essential services like healthcare and finance, their failure poses a systemic risk rather than a simple inconvenience. The author introduces a four-part threshold test to help leaders identify when AI has become indispensable to workforce capacity and public safety. Because AI often relies on complex, opaque chains of third-party providers, the source advocates for a governance model focused on resilience, continuity, and public-interest obligations. Ultimately, the piece calls for proactive mapping of AI dependencies to ensure that critical functions remain survivable even if the technology fails or is compromised.Full article: []

Narrated by NotebookLM