Use case #0001

How Risk Reporting AI assembles a CRO dashboard in under 60 seconds

A Chief Risk Officer who opens a risk dashboard at 8 AM should see the risk position as it stands at 8 AM — not as it stood when someone last ran the weekly report. The Risk Reporting Agent AI assembles the CRO dashboard in under 60 seconds from live data feeds across six operational systems, producing a single-view risk picture that reflects the portfolio's current state rather than its state at the last reporting cycle.

Why the CRO dashboard is almost always stale — and why that matters

A risk dashboard built from manually extracted, periodically refreshed data is almost always behind reality. The portfolio NPL figure shown on a Monday morning dashboard was computed on Friday — and it reflects the portfolio as it was on Thursday, because the Friday run used Thursday's CBS snapshot. An account that crossed the 90-day NPL threshold on Saturday is not on the Monday dashboard. A large LAP account that was written off on Friday afternoon is not on the Monday dashboard. The CRO is making risk-based decisions from a two-to-four day old picture of the portfolio.

In a stable, slowly changing portfolio, this lag is tolerable. In a portfolio where NPL classification runs daily and collections activity is continuous, a four-day dashboard lag is a governance failure. The CRO cannot perform their function — assessing whether the institution's risk position is within appetite, escalating emerging risks to the Board, directing management attention to deteriorating segments — from stale data.

The Risk Reporting Agent AI connects directly to the live data sources — the CBS, the Provisioning AI's classification database, the Stress Testing AI's scenario outputs, the collections system, and the Bureau pull log — and assembles the dashboard in under 60 seconds, on demand, showing the position as it stands at the moment of assembly.

"A risk officer whose dashboard is four days old is not managing risk — they are reviewing history. The Risk Reporting AI shows the present."

The data sources that feed the dashboard — and their latency

Core Banking System

Primary balance and payment data

Outstanding balances, payment status, GIRO / direct debit results, instalment schedule

CBS is the authoritative source for account balances and payment behaviour. The Risk Reporting AI pulls the live CBS balance for every NPL account and the payment register for the current day — not a nightly snapshot. Any GIRO / direct debit result received since midnight is included in today's dashboard.

Real-time
API pull

Provisioning AI

Classification and provision data

Current NPL classification, DPD count, provision required and held, 75-day watch list

The Provisioning AI runs its classification check at midnight every day and makes the results available via API. The CRO dashboard pulls the current classification for every account — post the overnight run — giving the CRO the day's freshest NPL picture before they open their laptop.

Daily
Post midnight

Stress Testing AI

Scenario output and CAR position

Latest stress scenario results, current CAR, rate shock impact at current portfolio composition

The most recent stress test scenario output is included in the CRO dashboard, showing the current CAR and the CAR under the MAS / Central Bank benchmark 200bps shock. If an on-demand stress test was run since the last dashboard assembly (triggered by an MAS / Central Bank rate change or CAR proximity alert), the latest result is used.

Latest run
Auto-refreshed

Collections System

Early bucket DPD and collection activity

DPD 1–90 distribution, collection contact rates, PTP rates, broken PTP counts, GIRO / direct debit failure rates

The collections system provides the leading indicators of future NPL: how many accounts are in the 60–89 DPD range (approaching the NPL threshold), what the current month's GIRO / direct debit failure rate is, and whether collection activity has resulted in PTPs at an acceptable rate. These are forward-looking credit quality signals.

Same-day
Intraday

Risk Pricing AI

NIM position and pricing metrics

Current portfolio NIM, yield by segment, rate-shock NIM compression, pricing adequacy vs risk

The Risk-Based Pricing AI's current NIM estimate is included in the dashboard, showing whether the portfolio is generating adequate margin relative to the risk being carried. The NIM figure is segment-level — the CRO can see which segments are generating adequate yield and which are not.

Current day
Market rates live

Origination AI

Pipeline quality and new disbursement risk

Current week disbursements, average credit score, TDSR (Total Debt Servicing Ratio) distribution, and score quality vs portfolio average

New disbursements create future risk — the CRO needs to see not just the current portfolio but the risk characteristics of what is being added to it. The Origination AI provides the current week's disbursement quality metrics: average credit score, TDSR (Total Debt Servicing Ratio) distribution, and product mix — so the CRO can assess whether origination quality is improving or deteriorating relative to the portfolio standard.

Daily
Current week

The live CRO dashboard: what assembles in 58 seconds

2.9% Gross NPL ratio ▲ +12bps vs last week · Watch

19.0% CAR (actual) 15.6% under 200bps stress

2.60% Portfolio NIM 1.48% under 200bps rate shock

84.2% Provision coverage ▼ Below 85% threshold · Alert

NPL distribution by segment — current

Home loan (salaried)

1.8% NPL — within appetite

1.8% ✓

SME business loan

3.8% NPL — approaching limit (4.5%)

3.8% ⚑

Personal loan (SE)

5.2% NPL — exceeds 4.5% appetite

5.2% ✗

LAP

3.1% NPL — within appetite

3.1% ✓

Gold loan

0.8% NPL — well within appetite

0.8% ✓

Leading indicators — early warning signals

Accounts 60–89 DPD

284 accounts · SGD38.4Cr · Up 18% vs last month

284 ⚑

GIRO / direct debit failure rate (Nov)

3.8% this month · vs 3.1% Oct average

3.8% ⚑

New disbursal avg credit bureau

718 this week · vs 712 portfolio avg · Improving

718 ✓

Active alerts requiring CRO attention

▲SE personal loan NPL at 5.2% — exceeds 4.5% risk appetite ceiling. Segment origination should be reviewed. 3 large accounts drive 41% of the segment NPL — account-level brief attached.

▲Provision coverage 84.2% — below 85% monitoring threshold. SGD24Cr additional provisioning required to restore coverage to threshold. Provisioning AI has flagged 18 accounts pending valuation update.

⚑60–89 DPD pool +18% vs last month — 284 accounts approaching NPL threshold. Collection team has contacted 71%. Remaining 29% (83 accounts) are uncontacted — escalation recommended.

✓CAR 19.0% — 4.0pp above 15% regulatory floor. Under 200bps stress CAR is 15.6%. Current buffer is adequate but management action underway (per Q3 stress test).

58sDashboard assembly time — 6 data sources, all live, assembled and formatted in 58 seconds on demand

2Critical alerts today — SE personal loan NPL above appetite and provision coverage below threshold

284Accounts in the 60–89 DPD pool — the forward-looking NPL indicator that predicts next quarter's classification run

ZeroManual steps between data sources and CRO dashboard — direct API pulls, no extract-transform-load

The 60-second dashboard is not a speed achievement — it is a governance requirement

A CRO who can see the current risk position in under a minute, on demand, any day of the week, is a CRO who is equipped to perform their function. A CRO who must wait for a weekly report or request a manual data extraction is structurally impaired in their ability to identify emerging risks before they become material events. The Risk Reporting Agent AI does not speed up the old process — it eliminates the wait by replacing the extraction-transformation-reporting chain with a direct connection between the live operational systems and the decision-maker who needs to see them. The 58 seconds is the connection time. The data has always been there.