Sovereign AI Strategy with Aki Ranin: Own Your Agents

Aki Ranin has spent years helping organizations think clearly about AI strategy, and right now he is focused on a problem most companies are only beginning to notice: the deeper you embed AI agents into your business, the more exposed you become to providers you do not control. In this episode, recorded soon after his keynote at apidays Singapore 2026, Aki walks through what sovereign AI actually means in practice — not as a regulatory concept, but as a question of operational risk. If your agents can be switched off without warning, if you cannot quantify hallucination rates, if you do not own the code, you do not own your business.What you will learn:Why the AI supply chain carries the same fragility risks as physical supply chains — from TSMC and ASML to the concentration of frontier models in Silicon ValleyHow a startup of 40 people running on Claude APIs was accidentally banned for 24 hours, and what that tells you about building on closed platformsThe Shopify case study: how one team reduced their AI costs from $5 million per year to $70,000 by moving down the model stack without sacrificing accuracyWhy "AI inflation" may matter as much as financial inflation — and how token costs are rising even as providers keep calling new models "Flash"The difference between drag-and-drop agent platforms and code-owned agents, and why that distinction will determine who keeps the IPHow to build a sovereign AI stack as a gradient: from paid cloud APIs, to private cloud instances, to open models, to on-premise inference hardwareWhy AI governance belongs at the board level, not delegated to a single AI leadHow Singapore's Foreign Minister Vivian Balakrishnan is running a local AI second-brain on a Raspberry Pi — and what that signals about accessibility.The sovereign AI stack Aki describes:Free tools — acceptable for non-sensitive, public-domain research onlyPaid subscriptions — minimum bar for any work touching company dataPrivate cloud instances (AWS, Azure, GCP) — frontier models with access controls and no training data exposureOpen models via API — more portability, less vendor lock-inOpen models on your own hardware — maximum auditability, task-specific agents, sensitive workloadsAbout Aki Ranin: Aki Ranin is a two-time founder, AI advisor, and published author who has focused exclusively on machine learning and AI since 2016. He co-created the "Usable AI" transformation framework, advises startups, corporates, and VC funds on AI strategy, and serves as Affiliate Faculty in AI and ML at Singapore Management University. Follow Aki Ranin on Substack, especially his article "Not your AI, not your business" — https://recursivelabs.substack.com/p/not-your-ai-not-your-business