This leading Singapore bank managed a commercial lending portfolio of over SGD 18 billion across Southeast Asia, serving mid-market enterprises and SMEs. Their credit risk assessment process relied on a combination of financial statement analysis, manual credit scoring models developed in the early 2010s, and relationship manager judgment. The average time to process a commercial loan application was 14 business days, with senior credit analysts spending approximately 6 hours per application on financial spreading, industry risk evaluation, and covenant structuring.
The bank's non-performing loan (NPL) ratio had crept up to 3.2% over the previous two years, above the industry median of 2.4% for Singapore-licensed banks. Post-mortem analysis of defaulted loans revealed that in 41% of cases, early warning signals were present in the borrower's financial data and industry trends but were missed during the manual review process. The credit committee acknowledged that their analysts simply could not process the volume and complexity of signals across a portfolio of over 4,200 active commercial borrowers.
Regulatory pressure from MAS was intensifying as well. New guidelines on technology risk management and model governance required banks to demonstrate that their risk models were robust, explainable, and regularly validated. The bank's existing models had not been fundamentally updated in over seven years.
Pertama Partners initiated the engagement with an AI Readiness Audit that mapped the bank's entire credit risk workflow from application intake through to portfolio monitoring. We analyzed the data infrastructure underlying their risk management process and found that the bank had rich datasets spread across five disconnected systems: core banking, financial spreading tools, external credit bureaus, industry databases, and relationship manager notes stored in an unstructured CRM.
During the AI Pilot Program, we developed an integrated AI risk assessment engine that combined traditional financial metrics with alternative data signals including real-time industry trend indicators, supply chain network analysis, and macroeconomic sensitivity modeling. The model was trained on eight years of the bank's own lending history encompassing over 12,000 loan outcomes, learning the nuanced patterns that distinguished performing loans from those that eventually deteriorated. We piloted the system with the bank's mid-market lending team across 200 new applications over a three-month period.
The AI Transformation Program then scaled the system bank-wide and established a comprehensive AI Governance framework. This included model documentation meeting MAS TRMG standards, quarterly model validation protocols, an explainability layer that generated natural language justifications for every risk recommendation, and a human-in-the-loop workflow where credit analysts reviewed and could override AI assessments with documented rationale.
"Pertama Partners helped us build a risk assessment capability that our credit analysts trust and our regulators respect. We are not just faster — we are fundamentally better at identifying risk before it materializes."— Tan Wei Lin, Chief Risk Officer
Every transformation starts with a conversation. Let's discuss your challenges and opportunities.
Discuss Your Challenge