AI investment is sky-high. ROI is zero. Only few have gone through this grind at enterprise scale and succeeded. I have.

Proving that an AI-driven operating model could work across fundamentally different business and support units, at a standard rigorous enough to justify enterprise-wide scaling.

Led flagship AI-driven operating model implementations across a transaction banking division and an internal audit function. Each implementation required building cross-functional workstreams from scratch, driving stakeholder alignment without formal authority, and codifying every learning into reusable playbooks.

The transaction banking division achieved 65% lead acceptance and 42% lift versus control. The internal audit function surfaced risks earlier and improved risk-type accuracy. Both became proof points that accelerated enterprise-wide scaling of the AI-driven operating model.

Private companies in the target segment had limited publicly available trade data, making it difficult to identify, qualify, and prioritise leads through traditional means. Relationship managers had no systematic way to assess which companies were worth pursuing or to build a compelling case for engagement before the first conversation.

Built a network graph AI model that used alternative signals — transaction patterns and corporate network connectivity — to identify and qualify high-potential prospects that traditional data sources could not surface. Rather than just generating a lead list, the model gave relationship managers contextual insights to make their conversations and proposals more relevant and compelling. Integrated into the corporate banking cross-sell engine at scale across six Asian markets.

SGD 6M incremental economic value in Year 1, scaling to SGD 10M in Year 2 across six markets. Model integrated into production and extended to additional use cases.

The FX business needed to grow revenue while improving customer satisfaction, in a highly competitive, margin-compressed environment.

Integrated AI-driven insights into sales, relationship manager, and back-end operations workflows and customer journey digital touchpoints. Re-organised the operating model into horizontal workstreams to break through silos and accelerate product and service time to market. Led delivery of cross-sell, attrition, and pricing propensity models.

15% FX revenue growth and 4.47/5 customer satisfaction score. This demonstrated that AI-driven personalisation can simultaneously improve commercial and customer outcomes.

As AI/ML adoption scaled, it became clear that existing model governance frameworks — designed for traditional statistical models — were not equipped to address the additional and emerging risks that AI/ML introduces. A new, dedicated governance approach was needed: to ensure all known and arising AI-specific risks were covered, and to formalise governance in a way that could scale with the organisation's growing AI/ML footprint.

Developed an interim AI governance checklist — grounded in direct experience auditing risk models under Basel, IFRS9, and MAS 637 — by bringing together everyone with a stake in AI governance: model developers, validators, technology, legal and compliance, and the business teams directly impacted. Co-developed a standardised AI delivery methodology, led the AutoML initiative, and built model inventory and explainability utilities aligned to MAS FEAT principles.

The interim governance checklist evolved into the enterprise's full AI governance framework, including a formal governance body. AI/ML model delivery time reduced from 4 months to 2 weeks, supporting hundreds of models in production.