Is OpenAI an Asian Story?
Key takeaways from the discussion of KewMann & ATP:
-
AI governance in regulated industries means more than accuracy; explainability and fairness are equally critical
-
The biggest challenge for banks is not building AI, it is defining where AI authority ends and human judgment begins
-
Even with 90-95% of loans automated, the cases escalated to humans are where real governance happens
-
Human-in-the-loop is a design principle from day one, not a fallback option
-
Risk teams are shifting from periodic reviews to continuous monitoring and predictive, prescriptive thinking
Episode’s agenda:
1. Softbank Is Making OpenAI Into a Uniquely Asian Story\
2. Softbank Building a Massive US Gas Power Plant
3. Welcome Yoke Ling Kew, Executive Director at KewMann to the Show (Click here to watch this session directly)
Frequently asked questions (FAQ)
Why is AI governance important for banks?
AI governance helps banks define how AI systems can use data, make recommendations and support decisions while maintaining accountability. This is particularly important in regulated areas such as lending, compliance, fraud detection and risk management, where explainability and human oversight may be required.
What does human-in-the-loop mean in AI governance?
Human-in-the-loop means people remain involved at important stages of an AI-driven process, particularly when a decision is complex, uncertain or high impact. The aim is not simply to review AI after something goes wrong, but to design human judgement into the workflow from the beginning.
Why are explainability and fairness important in financial AI?
Accuracy alone does not show whether an AI system is making decisions for appropriate reasons. Explainability helps organisations understand how outputs are produced, while fairness controls help identify whether decisions may create unintended or inappropriate differences across customers or groups.
Where should AI authority end and human judgement begin?
The boundary depends on the risk and consequence of the decision. Routine, low-risk processes may allow greater automation, while decisions involving regulatory interpretation, significant financial impact or unusual cases generally require stronger human oversight and clear escalation rules.
How is AI changing risk management in banks?
AI can help risk teams move from periodic reviews towards more continuous monitoring by analysing changing data and identifying emerging patterns earlier. Predictive models can indicate what may happen next, while more advanced AI systems can help prioritise or coordinate appropriate follow-up actions.
What role can Agentic AI play in regulated industries?
Agentic AI can coordinate multiple steps in a workflow, including gathering information, checking conditions, using different tools and initiating actions. In regulated environments, these capabilities require clear limits on authority, auditability, data governance and human approval for higher-risk decisions.
