UC5: Data Quality Exception Triage

The Problem

Banks operate data quality monitoring systems that flag anomalies in core banking, customer master, transaction, and risk data. High-volume environments generate 500-2000 alerts/day. Most (70-80%) are benign or self-remediating (e.g., overnight batch jobs creating temporary inconsistencies). Current manual triage involves data stewards reviewing each alert to classify severity and assign remediation. Median time to triage: 30-45 minutes per alert. Backlog of untriaged alerts reaches 5000+. Unaddressed data quality issues create downstream reporting errors and regulatory risk.

What the Agent Does

Receives data quality alert with affected entity (customer, account, transaction), issue type (null value, referential integrity, completeness, format, duplicate), and affected data domain
Queries data quality rules catalog to understand expected behavior
Applies known remediation patterns: 80% of exceptions follow 3-4 standard fixes (data refresh, batch rerun, master data correction, duplicate merge)
Routes simple exceptions to automated remediation workflow if confidence exceeds threshold (e.g., >90%)
Classifies alert by severity (critical/high/medium/low) based on impact on reporting or operations
Generates triage narrative (root cause assessment, recommended action)
Routes unclassified or novel exceptions to data steward queue for investigation
Tracks remediation outcome and updates pattern library with new findings

Data Requirements

Data Sources:

Data Quality Monitoring Platform (Informatica, Talend, Great Expectations, or custom DQ tool) , Real-time DQ alerts with entity, issue type, affected field
Core Banking System (Temenos T24, Finastra, Oracle FLEXCUBE) , Underlying data for quality checks, master data references
Data Warehouse/Snowflake , Transaction history, entity relationships for integrity validation
Data Quality Rules Catalog (internal) , Expected data patterns, remediation procedures for each issue type
ETL Logs (Talend, Apache Airflow) , Batch job execution logs for context when issues correlate to batch failures
Data Steward Knowledge Base (SharePoint, Confluence) , Common issues, remediation patterns, escalation criteria

Data Classification:

Public: None
Internal Structured: DQ alerts, alert metadata, rules catalog, remediation patterns, batch execution logs
Internal Unstructured: None
External Third-Party: None
Sensitive/Regulated: Affected customer/account data referenced in alerts (pseudonymized in DQ alerts only)

Data Quality Requirements:

Real-time alert ingestion (<1 minute latency from quality event to alert capture). Completeness: 99%+ of data quality issues detected (baseline per DQ monitoring system). Alert accuracy: 95%+ precision (false positive rate <5% after alert enrichment). Remediation pattern library freshness: updated monthly with new patterns discovered.

Integration Complexity: Medium , Requires API integration with DQ monitoring platform and core banking system. DQ rules catalog may be in separate PolicyTech platform. ETL logging may require custom queries. Automated remediation workflow (if enabled) requires write-access to master data systems, which carries operational risk and requires strict change controls.

Score Breakdown

Criterion	Weight	Score (1-5)	Weighted
Time Recaptured	15%	5	0.75
Error Reduction	10%	4	0.40
Cost Avoidance	10%	4	0.40
Strategic Leverage	5%	4	0.20
Data Availability	15%	5	0.75
Process Clarity	15%	4	0.60
Ease of Implementation	10%	3	0.30
Fallback Available	10%	5	0.50
Audience (Int/Ext)	10%	5	0.50
Composite	100%		4.30

Why It Scores Well

Alerts are structured (system-generated with consistent fields). Remediation rules are well-documented (80% of cases follow known patterns). High volume (500-2000 alerts/day) means massive time savings. Fallback is built-in: data stewards review anything the agent can't classify. Internal audience. Clear downstream value: faster data quality, better regulatory reporting, reduced incident escalations.

Regulatory Alignment

BCBS 239 (Principles for Effective Risk Data Aggregation): Agent supports rapid identification and remediation of data quality gaps
GLBA Safeguards Rule: Agent does not access customer PII; operates on data quality metadata only
Basel III Pillar 3: Agent ensures accuracy of regulatory reporting data

Sprint Factory Fit

Sprint 0 (2 weeks) + 3 build sprints (6 weeks)

Sprint 0: Data quality monitoring system integration, remediation rule codification, triage classification schema, escalation matrix definition

Build Sprints 1-3: Alert ingestion pipeline, pattern library development, remediation orchestration (for automated fixes), classification model training, steward escalation workflow

Comparable Implementations

Barclays: Uses ML-based triage for data quality alerts; 85% automated routing; 60% reduction in manual triage backlog
HSBC: Deployed exception pattern learning; 75% of alerts self-remediate within 2 hours

Deploy This Use Case with the Sprint Factory

From zero to a governed, production agent in 6 weeks.

Sprint Factory Schedule a Briefing

Related Use Cases

On-Demand Compliance

Regulatory Change Monitoring & Impact Triage

Score: 4.5/5.0

On-Demand Operations/IT

Internal IT Service Desk Triage & Routing

Score: 4.45/5.0

On-Demand Credit/Risk

Credit Risk Memo Drafting

Score: 4.4/5.0