With OECD now tracking 900+ AI governance frameworks across 18 jurisdictions, the industry has shifted from scarcity to abundance — making fragmentation and decision fatigue the central risk. Proposed treating AI governance as continuous telemetry-based monitoring (observability, controllability, stability, robustness, performance) and applying unified governance to both human and AI agents to avoid a "double standard".

Observed that in some leading firms, 30–40% of operational work is already being absorbed by digital twins of former employees — raising new questions around employee value assessment, privacy and dignity, knowledge ownership, workforce design and organisational trust. AI gives scalable intelligence; Web3 supplies scalable trust.

Deliberately inverted the usual framing — not what AI/Web3 can do for finance, but what finance can do for them. Discussions covered regulated stablecoins enabling Bitcoin neobanks that "rewrite payment unit economics"; Web3 as an immutable governance layer for AI ("since AI cannot be trusted to govern AI"); and Prof Lee's provocative formulation — that AI agents themselves should be made structurally inefficient to preserve human-centric outcomes.

Conducted in Mandarin, the panel extended AI's reach into specific professions: the classroom paradox where students use AI to write essays and teachers use AI to grade them — pushing the teacher's role upward on Bloom's taxonomy toward cultivating higher-order thinking (Chan); the unresolved question of who carries lifetime liability when AI does much of the engineering design (Zhou); AI's pressure on labour costs while virtual interaction may in fact increase demand for genuine face-to-face experiences (Shi); and AI enabling a shift from intuition-driven personalisation to data-driven consistency in retail healthcare (Wang). Dr Chua's closing reminder: as with road accidents, the underlying cause of most "AI harms" is human behaviour, not the machine.