Independent Advisor – Risk, Governance and AI Execution

A successful first Global Dialogue on AI Governance should move beyond principles and produce practical, execution-ready outcomes. First, it should establish a shared understanding that the primary gap is no longer technology, but execution — specifically in decision-making, ownership, and accountability within organizations. Second, the Dialogue should produce simple, actionable governance frameworks that can be applied at the point of execution, not only at the policy level. These should clarify who owns AI-driven decisions, how risks are managed in real time, and how accountability is enforced. Third, it should define measurable success criteria for AI initiatives, linking them to business outcomes, risk management, and public value — not just technical performance. Fourth, it should encourage cross-sector alignment by translating governance principles into operational practices that work consistently across public and private organizations. Finally, success would mean creating a practical roadmap for implementation, especially for emerging economies, where the challenge is not access to AI, but the capability to deploy it responsibly and effectively. In short, the Dialogue will succeed if it shifts the global conversation from "what AI should be" to "how AI is actually governed and executed in practice."

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

My selections reflect a practical, execution-focused perspective on AI governance. AI capacity-building is critical because many organizations do not lack access to AI, but lack the capability to deploy and manage it effectively. This includes not only technical skills, but also decision-making, risk management, and governance integration. Transparency, accountability, and human oversight are essential to ensure that AI systems remain controllable and aligned with organizational and societal objectives. However, these principles must be embedded into daily operations, not only defined at policy level. Safe, secure and trustworthy AI remains a foundation, particularly in managing risk, reliability, and public trust. Without this, large-scale deployment is not sustainable. Interoperability of governance approaches is increasingly important as organizations operate across jurisdictions and sectors. Governance frameworks must be adaptable and consistent enough to function in real-world, multi-stakeholder environments. Across all four priorities, the key challenge is not defining principles, but ensuring they are translated into consistent execution. This requires governance mechanisms that are practical, measurable, and integrated into actual decision-making processes. In that sense, effective AI governance is less about control structures, and more about enabling accountable, aligned, and scalable execution.

One critical cross-cutting issue not sufficiently captured is the "AI Execution Gap" - the gap between governance frameworks and actual decision-making in practice. Most current discussions focus on principles, policies, and high-level governance structures. However, in real organizational settings, failures often occur at the execution level: unclear ownership, fragmented decision flows, and governance that is reactive rather than embedded. This gap results in situations where AI systems are technically sound, but fail to deliver consistent, accountable, and scalable outcomes. Addressing this requires shifting focus toward "decision-centric governance" - ensuring that at every key decision point, there is clear ownership, defined accountability, and real-time oversight. Another emerging issue is the lack of integration between AI governance and operational workflows. Governance is often designed as a control layer, rather than being built into day-to-day processes. To move forward, global efforts should emphasize practical governance mechanisms that operate within real workflows, not outside them. Bridging the execution gap will be essential to ensuring that AI governance is not only well-designed, but actually effective in practice.

From my perspective in Asia and the Pacific, particularly in emerging economies, the most significant governance gap is not the absence of AI strategies or policies, but the lack of consistent execution at the organizational level. Many institutions have established high-level principles around trustworthy, ethical, and secure AI. However, these are often not translated into day-to-day decision-making processes. Governance remains policy-driven rather than operationally embedded. A key challenge is fragmented ownership. AI initiatives are frequently distributed across IT, data, and business units without clear accountability at the decision level. This results in slow implementation, duplicated efforts, and limited scalability. Another major gap is capability. While interest in AI adoption is high, there is insufficient integration of governance, risk, and business understanding. Organizations may invest in tools, but lack the internal capacity to align them with real business outcomes. Interoperability is also an emerging issue. As organizations operate across jurisdictions, inconsistent governance frameworks create complexity and uncertainty, particularly for cross-border collaboration. Despite these challenges, the opportunity is significant. Emerging markets have the advantage of leapfrogging legacy systems and adopting more integrated, execution-oriented governance models. If governance can be embedded into workflows — with clear ownership, aligned decision processes, and measurable accountability — AI can move from experimentation to scalable impact. In this context, the future of AI governance is not about more frameworks, but about making governance actionable, adaptive, and directly linked to execution outcomes.

The AI Dialogue can play a critical role by shifting international cooperation from high-level alignment to practical, execution-oriented collaboration. Today, many global AI governance efforts focus on principles and frameworks. While important, these often lack mechanisms for consistent implementation across different jurisdictions. The AI Dialogue can bridge this gap by promoting "operational interoperability" — not just aligning policies, but aligning how decisions are made and executed in practice. One key role is to facilitate shared models for governance-in-action. This includes common approaches to defining decision ownership, embedding accountability into workflows, and integrating governance into real operational processes. The Dialogue can also serve as a platform for exchanging practical use cases, where countries and organizations demonstrate how governance frameworks are applied in real environments — including successes and failures. Another important role is enabling capacity-building at the execution level, particularly in emerging economies. This goes beyond training on AI technologies to include governance design, risk management, and decision alignment. Finally, the AI Dialogue can help establish trust by promoting transparency in how AI systems are governed, not only how they are designed. In essence, its value lies in moving from alignment in theory to alignment in execution — enabling AI governance to function effectively across borders.

The AI Dialogue should build on and connect with existing global and regional initiatives such as the OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, and various national AI governance frameworks. These initiatives provide strong foundations in terms of principles, ethical standards, and policy direction. However, they often operate in parallel and at a high level, with limited integration at the operational layer. The added value of the AI Dialogue lies in acting as a "connector and translator" across these efforts. First, it can align these frameworks into practical, interoperable approaches that organizations can apply consistently across jurisdictions. This includes harmonizing governance concepts such as accountability, risk classification, and human oversight into usable models. Second, it can introduce execution-focused tools — such as decision flow mapping, ownership models, and governance checkpoints — that translate principles into action. Third, it can facilitate cross-sector collaboration by bringing together government, private sector, and technical communities to co-develop practical governance solutions. Finally, it can provide a feedback loop from implementation to policy, ensuring that real-world challenges inform future governance frameworks. In this way, the AI Dialogue does not replace existing initiatives, but enhances their impact by making them actionable, connected, and scalable in practice.

Different stakeholders should contribute based on their practical roles in the AI ecosystem, not only their formal positions. Governments can provide policy direction and regulatory frameworks, but should also share implementation challenges. The private sector can contribute real-world use cases and operational insights. Academia and the technical community can offer methodological rigor, while civil society ensures inclusiveness, ethics, and societal impact are addressed. To be effective, the AI Dialogue should move beyond a traditional conference format. I recommend a structured, multi-layered approach: First, focused thematic working groups that address specific governance challenges, such as decision accountability, risk management, and cross-border interoperability. Second, "execution labs" where participants work on real or simulated cases to translate principles into operational models. Third, structured synthesis sessions to consolidate insights into practical outputs — not just reports, but usable frameworks. Finally, a follow-up mechanism is essential. The Dialogue should not end with discussion, but continue through implementation tracking, peer learning, and iterative improvement. In this way, the AI Dialogue becomes not just a platform for exchange, but a system for collective problem-solving and continuous governance improvement.

Several important voices remain underrepresented in global AI governance discussions. First, practitioners at the operational level — those responsible for implementing AI systems within organizations. Their insights on execution challenges, decision-making constraints, and governance in practice are often missing. Second, small and medium-sized enterprises (SMEs) and organizations in emerging economies. These groups face different constraints compared to large institutions, particularly in terms of resources, capabilities, and access to expertise. Third, interdisciplinary professionals who bridge business, risk, and technology. AI governance is not only a technical or policy issue, but also an organizational and decision-making challenge. To include these voices, the AI Dialogue should actively design participation mechanisms beyond formal representation. This includes targeted outreach, simplified submission formats, and regional engagement channels. In addition, the Dialogue can create dedicated tracks for practitioners and SMEs, ensuring their experiences inform global discussions. Finally, translation — both linguistic and conceptual — is critical. Governance concepts must be made accessible and relevant across different contexts. Inclusion is not only about representation, but about ensuring that diverse perspectives meaningfully shape outcomes and implementation approaches.

To foster meaningful and dynamic engagement, the AI Dialogue should prioritize formats that are interactive, practical, and outcome-oriented. One effective format is "scenario-based workshops," where participants work through real-world governance challenges, such as deploying AI systems under regulatory constraints or managing cross-border data issues. Another is "execution sprints," short, focused sessions where stakeholders collaboratively design governance solutions, including decision flows, accountability models, and risk checkpoints. Peer-to-peer learning sessions are also valuable, allowing organizations to share practical experiences — both successes and failures — in a structured way. Digital collaboration platforms can extend engagement beyond the event itself, enabling continuous input, discussion, and refinement of ideas. Importantly, outputs should be concrete. Instead of general recommendations, the Dialogue should produce actionable tools, templates, and reference models that participants can immediately apply. Finally, feedback loops are essential. Participants should be able to report back on implementation progress, creating a cycle of learning and improvement. In this way, the AI Dialogue becomes a living process — not a one-time event — that continuously strengthens global AI governance in practice.

Effective AI governance requires approaches that translate principles into operational practice. One strong example is the use of risk-based frameworks, where AI systems are classified based on their potential impact, allowing governance efforts to be proportionate and focused. However, the effectiveness of such frameworks depends on how they are implemented within organizational processes. Another important practice is embedding governance directly into workflows. This includes defining clear decision ownership, establishing accountability checkpoints, and integrating governance into system design and deployment stages, rather than treating it as a separate compliance function. Model documentation and transparency tools, such as model cards and audit trails, also play a key role. These enable organizations to track how AI systems are developed, deployed, and monitored over time. In addition, cross-functional governance structures - bringing together business, technology, and risk teams - help ensure that AI decisions are aligned with organizational objectives and risk tolerance. From a practical perspective, platforms that support collaboration and continuous monitoring are essential. Governance should not be static, but adaptive, with mechanisms to update controls as systems evolve. Finally, capacity