01 / FEATURE
KYC & onboarding automation
ID extraction, sanctions screening, PEP review and adverse media — orchestrated with human gates at every high-risk step. Onboarding cycle measured in minutes, not days.
Compliance-first AI for finance.
MAS FEAT, IMDA, OJK and BNM alignment built in. Model risk managed, auditable, explainable. Built for banks, insurers, and capital markets across ASEAN.
The pressure
Why financial AI projects stall.
Most banks we meet have run a dozen pilots and shipped two. The blockers are rarely model accuracy — they’re governance, model risk, and the gap between data science and core banking systems.
Model risk that won’t sign off. MAS FEAT, model risk policy, and second-line review — every new model triggers a six-month evidence cycle the data team isn’t equipped to run.
Legacy core, modern ambition. Mainframe cores, fragmented customer data, and batch nightly cycles — fighting against real-time inference, streaming features, and continuous retraining.
False-positive fatigue. AML and fraud teams drowning in alerts. Investigators clearing 90%+ as noise — but the typology that matters is buried in the same pile.
Cross-border data, single architecture. PDPA in Singapore, OJK in Indonesia, BNM in Malaysia — each market wants residency, audit trails, and explainability with different evidence standards.
Where we ship
Use cases we’ve put into production.
Patterns we’ve shipped or are shipping today across ASEAN banks, insurers, and capital markets — each with measured accuracy targets and a governance evidence pack.
02 / FEATURE
AML transaction monitoring
Behavioural models on top of rules — cut false positives by 40–60% without missing typologies. Every alert ships with a reason code an investigator can defend.
03 / FEATURE
Credit underwriting co-pilots
Document extraction, financial-spread automation, and policy-aligned recommendations — the credit officer still owns the decision, the system does the prep.
04 / FEATURE
Fraud detection at the edge
Card and digital-channel fraud scored in single-digit milliseconds. Drift monitored continuously; new typologies retrained weekly, not quarterly.
05 / FEATURE
Customer-service triage
Retail and priority-banking enquiries routed by intent and sentiment. Grounded answers from product docs, T&Cs and account context — never invented.
06 / FEATURE
Regulatory reporting copilots
MAS, OJK and BNM submissions assembled from internal data with full lineage. Auditor-ready evidence ships with every report, not after the fact.
Real-world example
From 41 pilots to six in nine months.
An ASEAN universal bank had run 41 AI pilots over three years. Two were live. We ran an eight-week strategy engagement under MAS FEAT and the bank’s own model risk policy — closed 35, consolidated six into a delivery roadmap, and shipped three to production within nine months.
Before
Status quo
- 41 active pilots, 2 in production
- No common model risk template
- Six-month average pilot-to-production cycle
- Each business unit running its own MLOps stack
After
Post-engagement
- Six prioritised use cases, three in production
- One MRM template approved by second line
- Eight-week production cycle as the new standard
- Single platform, paved-road deployment
ASEAN universal bank · post-engagement
Solutions that fit
Where to start, by maturity.
Strategy & Governance → Set the AI portfolio direction and the model risk operating model.
AI Sprint — 4 weeks → Validate a use case end-to-end with a working prototype and feasibility report.
Accelerate — embedded → One senior engineer in your sprint cadence, 3–6 months, monthly cancel.
Deploy — production SLA → Two to three engineers, full FORGE, monitoring and SLA-backed support.
Compliance & assurance
Frameworks we build against.
Governance is not a review gate at the end — it’s the constraint set the system is designed to satisfy. Every engagement ships with the evidence pack your second line and external auditors expect.
Frameworks we build against
MAS FEAT, MRM, and ASEAN data residency — instrumented, not bolted on.
MAS FEAT. Fairness, Ethics, Accountability, Transparency get instrumented into the model lifecycle, not bolted on. The evidence pack maps directly to MAS principles.
Model Risk Management. Aligned to MAS Notice 612 and its equivalents across ASEAN. Independent validation, ongoing monitoring, and challenger models built into the operating model.
PDPA and cross-border. Singapore PDPA, Indonesia OJK and Malaysia BNM data residency get designed in. Tenant isolation by default; nothing leaves the jurisdiction without explicit consent.
Have an AI portfolio that won’t ship?
Book a 30-minute call with our financial services lead. We’ll bring two named references and the evidence pack template.
Talk to financial services lead
30 minutes · reply within 1 business day