Q1. Could you start by giving us a brief overview of your professional background, particularly focusing on your expertise in the industry?

I have spent over 25 years navigating the intersection of banking, technology, and consulting. My career has essentially been a journey from traditional operations at HSBC and Bank of America to architecting the 'AI-first' wholesale bank. 

Currently, I lead AI Transformation for Corporate Banking in SEA in a management consulting capacity, where I’m spearheading 'Program AURA - transitioning a Tier-1 bank to an autonomous wholesale credit model. 

Before this, at First Abu Dhabi Bank (FAB), I led the deployment of Agentic AI workflows for trade finance that didn't just 'assist' but achieved 45% efficiency gains and 90%+ accuracy in complex document screening. My value proposition is that I don't just talk about the tech; I’ve managed the $10M+ P&Ls and the regulatory scrutiny that comes with it.

Q2. In a market environment increasingly defined by unpredictable and high-impact geopolitical shifts, how are autonomous risk models being calibrated to account for non-linear events that 'clean data' history cannot predict? What is the industry standard for 'human-in-the-loop' intervention in AI-led trading and treasury?

My perspective is grounded in treasury enablement and AI-led wholesale banking. We have to move past the idea that 'clean data' history is a crystal ball; it isn't. In my experience, the industry standard isn't about building a perfect predictor, but a rapid-response framework. The approach I advocate combines AI models with dynamic stress overlays and real-time event triggers. On 'human-in-the-loop,' the trend is 'Bounded Autonomy'. Humans define the risk appetite and 'kill switches,' while the AI operates the engine within those guardrails. If an event falls outside the preset policy—such as a sudden geopolitical shock—the system must automatically 'escalate to human' rather than attempt to guess.

Q3. How is the adoption of real-time, AI-driven AR/AP forecasting impacting the systemic requirement for liquidity buffers? Are we seeing a trend where banks with superior predictive capabilities can operate with lower capital reserves, and how is that surplus being redeployed?

We are seeing a massive shift. At FAB, we monetized predictive analytics for corporate clients, creating new revenue streams in cash forecasting that saw a 12% fee uplift in just four months. Better forecasting certainly reduces 'idle' operational balances and improves working-capital deployment. However, it’s a mistake to conflate operating liquidity with regulatory capital. 

While AI allows banks to operate more leanly day-to-day, the surplus is usually redeployed into more aggressive product pricing or targeted treasury yields rather than just 'lowering the buffer' below prudential requirements. It’s about balance sheet efficiency, not regulatory evasion.

Q4. With the B2B embedded finance market projected to hit $15.6 trillion by 2030, how are institutions moving beyond simple 'payment processing' to offer high-margin credit and working capital directly within third-party SaaS and ERP platforms?

The market is moving from embedding 'payments' to embedding 'financial workflows.' Payments are now a commodity. The high-margin opportunity lies in receivables, reconciliation, and supplier finance directly within the ERP. Have seen this play out first-hand. 

At Accenture, I helped build a multi-party blockchain platform for Siam Commercial Bank, delivering 33% cost savings and a 35% increase in Supply Chain Finance (SCF) leads. To win here, institutions need to move beyond APIs and start embedding their underwriting logic into the customer’s daily software. That is how you capture the $15.6 trillion opportunity.

Q5. With the Central Bank of the UAE’s 2026 Guidance Note on AI, how are institutions solving the 'explainability gap' in autonomous nudge engines to ensure they remain compliant while operating without human-in-the-loop oversight?

It is essential to follow the CBUAE guidance closely. The 'explainability gap' is often a 'governance gap. We solve this by ensuring nudge engines are never a 'black box.' Every output must have an audit trail and a reason code. For high-stakes decisions - like a credit nudge - the engine must be able to 'show its work' against documented policy constraints. If the system can’t explain why it recommended a specific limit or price, it shouldn't be live in a regulated environment like the UAE.

Q6. As we move from Chatbots to Agentic AI, what safety measures and identity protocols are in place to stop a machine error from spiraling into a systemic financial loss?

The risk with Agentic AI is Authority, not just Intelligence. The moment an agent can 'execute,' you need Identity Protocols for machines. I advocate for a 'Least-Privilege' architecture. No single agent should have the right to initiate, approve, and reconcile a transaction. We apply traditional 'maker-checker' controls to AI agents. By separating 'read' rights from 'execute' rights and implementing hard transaction caps, we ensure that a machine error is contained before it can ever become a systemic loss.

Q7. If you were an investor looking at companies within the space, what critical question would you pose to their senior management?

My key question would be: 'Is your growth based on a better algorithm, or a better moat?' In this space, an algorithm can be copied. A moat is built through privileged workflow data, defensible distribution into ERPs, and a regulatory architecture that can withstand a downturn. I want to see a management team that isn't just chasing AI hype but has built a model where the economics still work after accounting for the 'cost of control' and regulatory overhead.