Financial institutions are moving well beyond experimentation with AI, deploying it across credit decisioning, fraud detection, and everyday employee workflows at genuine scale. But how they get there differs sharply: some are building their own foundation models on proprietary data to avoid depending on external providers with their most sensitive information; others adopt commercial AI tools but wrap them in anonymization and strict governance before sensitive data reaches them; and some capital allocators are treating AI infrastructure itself — compute, data centres, foundation models — as a strategic asset worth direct ownership.

Regulation is catching up fast and unevenly, with major regimes now treating data location and third-party AI access as a risk to be supervised, not just a technology choice. Given the different AI strategies already in motion, what would it actually take to make them interoperable, auditable, and defensible across jurisdictions?

Agentic commerce moved from feature to production in barely eighteen months - from AI-assisted purchasing in early 2025 to the first production-grade agentic transaction run on full banking rails in June 2026. But the infrastructure underneath is still fragmented: at least half a dozen competing protocols govern identity, authorization, and payment - and the platforms actually processing agent transactions haven't converged on any of them.

From the standards that can bind an agent to accountable identity, the platforms that process its transactions, and the frameworks that would assign liability when one goes wrong, this roundtable convenes financial institutions, payment processors, standard-setting bodies, and compliance and enforcement bodies to discuss what each party would need to adopt, require, or share to close the trust gap, and on what timeline.

The discussion will use concrete examples to test and refine trust principles for agentic commerce, including what evidence would demonstrate them in practice, what should be tested or benchmarked before an agent is trusted to transact, and what different actors would need to recognise across protocols and jurisdictions.

The next phase of AI in finance will not be defined by better models alone, but by institutions that can successfully embed AI into regulated business processes at scale. From compliance and risk management to customer servicing and agentic finance, AI is creating new opportunities to transform how financial services are designed, delivered and supervised. As institutions explore frontier AI capabilities, the challenge is no longer just building AI solutions, but operationalising them responsibly while maintaining trust, resilience and accountability.
 
Many financial institutions face similar challenges in developing, evaluating and deploying AI, yet often tackle them independently. Through BuildFin.ai, financial institutions, technology providers and research institutes are collaborating to develop and scale frontier AI capabilities within established regulatory and risk management frameworks. This includes the development of the MindForge AI Risk Management Toolkit, which provides practical guidance, industry case studies and implementation insights, as well as the SAFR (Safeguards for Agentic Finance at Runtime) framework, which explores safeguards for AI-enabled financial activities. This roundtable brings together members of the BuildFin.ai community and industry leaders to exchange lessons from the front lines of AI deployment and discuss how collaboration can help accelerate the responsible adoption of AI across financial services.