Executive Leadership Journey: Strategic Risk and Growth Analytics
Persona: Elena, Chief Risk Officer (CRO) at Meridian Bank (20 years in financial services, reports to CEO and Board) Objective: Use the platform for strategic portfolio analysis, regulatory oversight, and board-level decision support -- without writing SQL or navigating technical interfaces Primary interface: Agentic Workbench (natural language) and BI Dashboards (visual) Datasets: All Meridian Bank datasets (see Industry Overview)
Stage 1: Ingestion
Elena does not configure data sources directly. Her team -- including Amir (data science), Kenji (ML engineering), and Rachel (BI/regulatory) -- has connected all enterprise data sources. Elena's engagement with ingestion is through visibility and oversight.
Data Landscape Overview
In the Agentic Workbench, Elena asks:
Elena: "What data sources are connected to the platform and when was each last synced?"
The AI agent queries the catalog and returns:
| Source | Type | Tables | Last Sync | Freshness SLA | Status |
|---|---|---|---|---|---|
| Core Banking | PostgreSQL CDC | 12 | 14 min ago | 15 min | On time |
| Risk Systems | PostgreSQL | 8 | 2h 12m ago | Daily | On time |
| Bloomberg Market Data | REST API | 3 | 18h ago | Daily (after close) | On time |
| Snowflake Analytics | Snowflake | 6 | 5h ago | Daily | On time |
| Bureau Scores | CSV import | 1 | 12 days ago | Monthly | On time |
| SWIFT Payments | Kafka stream | 1 | 1.2s ago | Real-time | On time |
Elena: "Good. Are there any data quality issues I should know about?"
Agent: "Two active warnings: (1)
credit_applications.income_reportedhas 7.9% missing values -- the data science team is aware and imputes these in their models. (2) Three instruments inmarket_datahave stale pricing (last updated 3 days ago) -- flagged to the market data team. No critical issues affecting regulatory reporting."
Stage 2: Discovery
Elena uses natural language to explore the data landscape relevant to her strategic questions.
Exploring the Commercial Lending Portfolio
Elena: "What data do we have on our commercial lending portfolio?"
The agent searches the catalog and returns a structured summary:
Commercial Lending Data Assets
================================
Tables found: 8 across 3 schemas
lending.commercial_loans 12,841 rows Quality: 98.2%
lending.commercial_facilities 4,219 rows Quality: 99.1%
lending.covenant_tracking 38,412 rows Quality: 97.8%
risk_analytics.commercial_pd 12,841 rows Quality: 99.5%
risk_analytics.commercial_lgd 12,841 rows Quality: 99.3%
regulatory_output.large_exposures 241 rows Quality: 100%
market_data.sector_indices 1,825 rows Quality: 100%
historical.commercial_defaults 2,104 rows Quality: 99.8%
Key metrics available via semantic layer:
- Commercial NPL ratio
- Sector concentration index
- Covenant breach rate
- Weighted average PD
- Expected credit loss (ECL)
Data lineage: Full traceability from source systems through
risk models to regulatory outputs.Elena: "What is the quality score methodology?"
Agent: "Quality scores are computed by the Data Quality Service using Great Expectations. They measure completeness (no nulls in required fields), validity (values within expected ranges), consistency (cross-table referential integrity), and timeliness (data freshness vs SLA). A score of 98.2% means 1.8% of quality checks had warnings -- typically minor issues like optional fields left blank. No critical failures in any commercial lending table."
Stage 3: Query
Elena asks business questions in natural language. The AI agent generates SQL, executes it, and returns results with visualizations.
Portfolio Exposure Analysis
Elena: "What is our current exposure to the real estate sector and how has it changed in the last 12 months?"
The agent generates and executes a federated query across credit and market data:
-- Generated by AI Agent for Elena's query
-- Sources: lending.commercial_loans (PostgreSQL) + market_data.sector_indices (Bloomberg)
WITH current_exposure AS (
SELECT
sub_sector,
COUNT(*) AS num_facilities,
SUM(committed_amount) AS total_committed,
SUM(outstanding_balance) AS total_outstanding,
AVG(pd_score) AS weighted_avg_pd,
SUM(outstanding_balance) / SUM(SUM(outstanding_balance)) OVER() * 100
AS pct_of_portfolio
FROM lending.commercial_loans cl
JOIN risk_analytics.commercial_pd cpd ON cl.loan_id = cpd.loan_id
WHERE cl.sector = 'Real Estate'
AND cl.status = 'ACTIVE'
GROUP BY sub_sector
),
historical_trend AS (
SELECT
DATE_TRUNC('month', reporting_date) AS month,
SUM(outstanding_balance) AS total_outstanding
FROM lending.commercial_loans
WHERE sector = 'Real Estate'
AND reporting_date >= CURRENT_DATE - INTERVAL '12 months'
GROUP BY DATE_TRUNC('month', reporting_date)
)
SELECT * FROM current_exposure
ORDER BY total_outstanding DESC;The agent presents the results:
| Sub-Sector | Facilities | Committed ($M) | Outstanding ($M) | % Portfolio | Avg PD |
|---|---|---|---|---|---|
| Commercial Office | 142 | $1,240 | $980 | 8.2% | 3.1% |
| Residential Development | 89 | $890 | $720 | 6.0% | 2.4% |
| Retail Properties | 67 | $520 | $410 | 3.4% | 4.2% |
| Industrial/Logistics | 54 | $480 | $390 | 3.3% | 1.8% |
| Mixed Use | 31 | $290 | $230 | 1.9% | 2.7% |
| Total Real Estate | 383 | $3,420 | $2,730 | 22.8% | 2.8% |
Agent: "Real estate represents 22.8% of the commercial portfolio, up from 19.4% twelve months ago. The increase is driven primarily by Commercial Office (+120M). Note that Commercial Office has the second-highest average PD at 3.1%, and Retail Properties is showing elevated risk at 4.2% PD. The concentration is approaching the 25% internal limit set by the Board Risk Committee."
Elena: "Show me the 12-month trend."
The agent generates a trend chart showing monthly outstanding balances by sub-sector, highlighting the upward trajectory in Commercial Office.
Stage 4: Orchestration
Elena receives automated briefings and can request ad-hoc analyses through the platform.
Weekly Risk Briefing
Every Monday at 8:00 AM, Elena receives an automated risk briefing assembled by the platform:
┌─────────────────────────────────────────────────────────────┐
│ WEEKLY RISK BRIEFING │
│ Week of February 24, 2026 │
├─────────────────────────────────────────────────────────────┤
│ │
│ CAPITAL POSITION │
│ CET1 Ratio: 12.3% (target: >10.5%) [ON TARGET] │
│ Change vs last week: -0.1pp │
│ Driver: RWA increase from new commercial originations │
│ │
│ CREDIT QUALITY │
│ NPL Ratio: 2.1% (target: <3.0%) [ON TARGET] │
│ New NPLs this week: 3 (total: $4.2M) │
│ Watchlist additions: 7 accounts │
│ │
│ FRAUD OPERATIONS │
│ Fraud catch rate: 95.1% (target: >95%) [ON TARGET] │
│ False positive rate: 1.6% (target: <2%) [ON TARGET] │
│ Revenue protected this week: $1.42M │
│ │
│ LIQUIDITY │
│ LCR: 138.7% (minimum: >100%) [ON TARGET] │
│ NSFR: 112.4% (minimum: >100%) [ON TARGET] │
│ │
│ CONCENTRATION ALERTS │
│ [WARNING] Real estate: 22.8% (limit: 25%) │
│ [OK] Top 10 borrowers: 14.2% (limit: 20%) │
│ [OK] Country risk: all within limits │
│ │
│ REGULATORY CALENDAR │
│ Basel III monthly report: due March 10 (8 business days) │
│ Quarterly stress test: due March 31 (21 business days) │
│ │
└─────────────────────────────────────────────────────────────┘Requesting Ad-Hoc Analysis
Elena: "I need a stress test scenario for our commercial real estate portfolio. Can you set that up as a workflow for the data team?"
Agent: "I will create a workflow request. What scenarios would you like analyzed?"
Elena: "Two scenarios: (1) commercial real estate values decline 20% with vacancy rates increasing to 15%, and (2) interest rates increase 200 basis points with a mild recession."
The agent creates a structured workflow request:
{
"workflow_request": {
"title": "CRE Portfolio Stress Test - CRO Request",
"requestor": "elena@meridian.bank",
"priority": "high",
"deadline": "2026-03-07",
"scenarios": [
{
"name": "CRE Downturn",
"parameters": {
"property_value_decline": -0.20,
"vacancy_rate_increase": 0.15,
"rental_income_decline": -0.12,
"cap_rate_expansion": 0.015
}
},
{
"name": "Rate Shock + Recession",
"parameters": {
"interest_rate_change_bps": 200,
"gdp_growth": -0.015,
"unemployment_increase": 0.025,
"credit_spread_widening_bps": 150
}
}
],
"outputs_requested": [
"Portfolio loss projection by sub-sector",
"Impact on CET1 ratio",
"List of facilities likely to breach covenants",
"Comparison to current provisions"
],
"assigned_to": ["amir@meridian.bank", "rachel@meridian.bank"]
}
}Stage 5: Analysis
Elena asks analytical questions and receives AI-powered analysis with statistical rigor.
Early Warning Analysis
Elena: "Which loan segments are showing early warning signs?"
The agent runs a delinquency trend analysis with statistical significance testing:
-- Agent-generated: Early warning signal detection
-- Identifies segments with statistically significant deterioration
WITH segment_trends AS (
SELECT
sector,
loan_product,
DATE_TRUNC('month', reporting_date) AS month,
COUNT(*) AS total_loans,
SUM(CASE WHEN days_past_due > 0 THEN 1 ELSE 0 END) AS delinquent,
AVG(pd_score) AS avg_pd
FROM lending.commercial_loans cl
JOIN risk_analytics.commercial_pd cpd ON cl.loan_id = cpd.loan_id
WHERE reporting_date >= CURRENT_DATE - INTERVAL '6 months'
GROUP BY sector, loan_product, DATE_TRUNC('month', reporting_date)
)
SELECT
sector,
loan_product,
-- Current vs 6-month-ago comparison
MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE) THEN avg_pd END) AS current_pd,
MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE - INTERVAL '6 months') THEN avg_pd END) AS pd_6m_ago,
MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE) THEN avg_pd END)
- MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE - INTERVAL '6 months') THEN avg_pd END)
AS pd_change
FROM segment_trends
GROUP BY sector, loan_product
HAVING MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE) THEN avg_pd END)
> MAX(CASE WHEN month = DATE_TRUNC('month', CURRENT_DATE - INTERVAL '6 months') THEN avg_pd END)
ORDER BY pd_change DESC;Agent: "Three segments are showing statistically significant deterioration (p < 0.05):"
| Segment | Current PD | 6M Ago PD | Change | Significance | Exposure ($M) | Action Recommended |
|---|---|---|---|---|---|---|
| Retail Properties / Term Loans | 4.2% | 2.8% | +1.4pp | p = 0.003 | $310 | Enhanced monitoring |
| Commercial Office / Construction | 3.8% | 2.9% | +0.9pp | p = 0.012 | $180 | Covenant review |
| Hospitality / Revolving Credit | 3.1% | 2.4% | +0.7pp | p = 0.031 | $95 | Watchlist screening |
Agent: "The retail properties segment is the most concerning -- PD has increased 50% in 6 months. This correlates with a 12% decline in foot traffic at malls in our geographic footprint (source: market_data.sector_indices). I recommend requesting a detailed portfolio review from the credit committee."
Elena: "What is the total exposure to these three segments and what would a 2x default rate mean for our provisions?"
Agent: "Total exposure: 18.4M to $24.1M (depending on LGD assumptions). This would reduce the CET1 ratio by approximately 0.15 percentage points -- from 12.3% to 12.15%, still well above the 10.5% target."
Stage 6: Productionization
Elena subscribes to dashboards and automated reports for ongoing decision support.
CRO Dashboard Subscription
Elena's personalized dashboard in the BI Workbench shows the metrics most relevant to her role:
┌─────────────────────────────────────────────────────────────────┐
│ CRO EXECUTIVE DASHBOARD │
│ Elena Vasquez | Updated: Real-time │
├──────────────────────────────────┬──────────────────────────────┤
│ │ │
│ Portfolio Health Score │ Capital Position │
│ │ │
│ ████████████░░ 82/100 │ CET1: 12.3% ▲ +0.2pp YTD│
│ │ T1: 14.1% ▲ +0.1pp │
│ vs last month: +3 │ TCR: 16.8% ▲ +0.3pp │
│ vs last year: +8 │ │
│ │ Buffer over minimum: │
│ │ ████████████████ 7.8pp │
├──────────────────────────────────┼──────────────────────────────┤
│ │ │
│ Concentration Limits │ Model Performance │
│ │ │
│ Real Estate ████████████░ 23% │ Credit PD AUC: 0.84 │
│ Top 10 ██████░░░░░░ 14% │ Fraud Det AUC: 0.97 │
│ Country ████░░░░░░░░ 8% │ LGD Model AUC: 0.71 │
│ │ │
│ [!] RE approaching 25% limit │ All models within │
│ │ validation thresholds │
├──────────────────────────────────┴──────────────────────────────┤
│ Key Risk Indicators (trend: 12 months) │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ NPL Ratio Fraud Loss Rate VaR │ │
│ │ 3%│ 0.05%│ 2.5%│ │ │
│ │ │ ╭──╮ │ │ │ │
│ │ 2%│──╯ ╰── │ ──────────── 1.8%│──── │ │
│ │ │ ──── 2.1% │ │ │ │
│ │ 1%│ 0.02%│──────────── │ │ │
│ │ └────────────── └────────────── └──── │ │
│ │ M A M J J A S O N D J F (months) │ │
│ └──────────────────────────────────────────────────────────┘ │
└────────────────────────────────────────────────────────────────┘Board-Ready Report Delivery
Elena subscribes to automated report delivery for board meetings:
| Report | Frequency | Delivery | Format | Audience |
|---|---|---|---|---|
| CRO Risk Summary | Weekly (Monday 7am) | Email + PDF | 2-page executive brief | Elena |
| Board Risk Report | Monthly (before board meeting) | Secure portal + PDF | 15-page comprehensive | Board Risk Committee |
| Regulatory Dashboard Snapshot | Daily (8am) | HTML summary with links | Elena + CFO | |
| Model Validation Summary | Quarterly | Email + PDF | 5-page validation report | Elena + Model Risk team |
Stage 7: Feedback
Elena receives real-time alerts and periodic digests to stay informed without information overload.
Weekly KPI Digest
Every Monday morning, Elena receives a concise digest:
From: MATIH Platform <alerts@matih.meridian.internal>
To: Elena Vasquez <elena@meridian.bank>
Subject: Weekly Risk KPI Digest - Week of Feb 24, 2026
SUMMARY: 11 of 12 KPIs on target. 1 warning.
KPI Current Target Status
─────────────────────────────────────────────────────────
CET1 Ratio 12.3% >10.5% OK
NPL Ratio 2.1% <3.0% OK
Fraud Catch Rate 95.1% >95% OK
False Positive Rate 1.6% <2% OK
LCR 138.7% >100% OK
NSFR 112.4% >100% OK
Portfolio VaR 1.8% <2.5% OK
Credit Loss Rate 1.05% <1.2% OK
Regulatory Report Accuracy 100% 100% OK
Customer Acquisition Cost $412 <$350 MISS
RE Concentration 22.8% <25% WARN
Model AUC (Credit PD) 0.84 >0.75 OK
ACTION ITEMS:
[WARN] Real estate concentration at 22.8%, approaching
25% limit. Trend: +0.6pp in 3 months.
Recommendation: Review new origination pipeline.
[MISS] Customer acquisition cost $412 vs $350 target.
Trend: stable. Marketing team reviewing channel mix.Real-Time Concentration Limit Alerts
{
"alert": {
"name": "concentration-limit-warning",
"type": "threshold",
"metric": "sector_concentration_pct",
"conditions": [
{
"level": "warning",
"threshold": "limit * 0.90",
"message": "Sector concentration at 90% of limit"
},
{
"level": "critical",
"threshold": "limit * 0.95",
"message": "Sector concentration at 95% of limit -- origination review required"
},
{
"level": "breach",
"threshold": "limit * 1.00",
"message": "LIMIT BREACH -- immediate escalation to Board Risk Committee"
}
],
"channels": {
"warning": ["email:elena@meridian.bank"],
"critical": ["email:elena@meridian.bank", "sms:elena", "slack:#risk-escalation"],
"breach": ["pagerduty:board-risk", "email:board-risk-committee", "sms:elena"]
}
}
}Regulatory Deadline Reminders
| Days Until Deadline | Action | Notification |
|---|---|---|
| 10 | Pipeline starts automatically | Slack notification to Rachel |
| 5 | Status check: data quality, calculation complete | Email to Elena + Rachel |
| 3 | Approval gate: report requires sign-off | Email + dashboard alert to Elena + CFO |
| 1 | Final review: report ready for submission | SMS to Elena if not yet approved |
| 0 | Submission deadline | Automated submission if approved; emergency escalation if not |
Stage 8: Experimentation
Elena uses the Agentic Workbench to run scenario analyses and compare strategic options.
Scenario Modeling
Elena: "Model two scenarios for me: interest rate increase of 200 basis points, and a recession scenario. Compare the portfolio loss projections for each."
The agent runs the analysis:
┌─────────────────────────────────────────────────────────────────┐
│ SCENARIO COMPARISON │
│ Meridian Bank CRE Portfolio │
├──────────────────────┬──────────────────────┬───────────────────┤
│ │ Scenario 1: │ Scenario 2: │
│ Metric │ Rate +200bps │ Recession │
├──────────────────────┼──────────────────────┼───────────────────┤
│ Portfolio PD (wt avg) │ 3.4% (from 2.8%) │ 5.1% (from 2.8%) │
│ Expected Loss │ $42M (from $31M) │ $68M (from $31M) │
│ NPL Ratio │ 2.8% (from 2.1%) │ 4.2% (from 2.1%) │
│ CET1 Ratio Impact │ 12.0% (from 12.3%) │ 11.4% (from 12.3%)│
│ Provisions Required │ +$11M │ +$37M │
│ Covenant Breaches │ 12 facilities │ 31 facilities │
│ Capital Buffer │ 7.5pp above min │ 6.9pp above min │
├──────────────────────┼──────────────────────┼───────────────────┤
│ Risk Assessment │ Manageable │ Requires action │
└──────────────────────┴──────────────────────┴───────────────────┘Agent: "Key findings:"
"Scenario 1 (Rate +200bps): Moderate impact. CET1 ratio remains well above target at 12.0%. The primary risk is to floating-rate commercial real estate loans where debt service coverage ratios could deteriorate. 12 facilities would likely breach interest coverage covenants. Recommended action: proactive outreach to these borrowers."
"Scenario 2 (Recession): Significant impact. CET1 drops to 11.4% -- still above the 10.5% target but with reduced buffer. NPL ratio would exceed the 3% internal target. 31 covenant breaches expected, concentrated in retail properties (14) and hospitality (9). Additional provisions of $37M would be required. Recommended action: review origination appetite for high-risk sectors, accelerate collection efforts on watchlist accounts."
"Confidence intervals: These projections use historical loss data from the 2008-2009 and 2020 downturns, calibrated to the current portfolio composition. The 90% confidence interval for recession losses is 89M."
Elena: "What if both scenarios happen simultaneously?"
Agent: "A combined scenario (rate shock + recession) would produce an estimated expected loss of $94M, with CET1 declining to 10.8%. This would breach the internal 10.5% target buffer, though it remains above the regulatory minimum of 4.5%. I recommend presenting this combined scenario to the Board Risk Committee as a tail risk assessment. Shall I generate a board-ready presentation with these findings?"
Elena: "Yes, format it for the March board meeting."
The agent generates a structured board presentation with executive summary, scenario details, risk heat maps, recommended actions, and appendices -- delivered as a versioned PDF to Elena's secure document portal.
Key Takeaways
| Stage | Key Action | Platform Component |
|---|---|---|
| Ingestion | Viewed data landscape and quality status via natural language | Agentic Workbench, Catalog Service |
| Discovery | Explored commercial lending data assets, reviewed quality scores | Agentic Workbench, Data Quality Service |
| Query | Asked natural language questions, received federated SQL results | Agentic Workbench, Text-to-SQL, Query Engine |
| Orchestration | Received automated weekly briefings, requested ad-hoc stress tests | Pipeline Service, workflow automation |
| Analysis | Early warning detection with statistical significance testing | AI Agent, Data Quality Service |
| Productionization | Subscribed to CRO dashboard, board-ready automated reports | BI Workbench, BI Service |
| Feedback | Weekly KPI digests, real-time concentration alerts, deadline reminders | Alerting, monitoring |
| Experimentation | Multi-scenario portfolio stress testing with confidence intervals | Agentic Workbench, AI Service |
Related Walkthroughs
- Data Scientist Journey: Credit Risk Scoring -- Amir builds the PD models that feed Elena's portfolio analytics
- ML Engineer Journey: Real-Time Fraud Detection -- Kenji operates the fraud system whose KPIs Elena monitors
- BI Lead Journey: Regulatory Reporting -- Rachel builds the dashboards and reports Elena consumes
- Financial Services Overview -- Industry context and sample datasets