Overview
The remediation agent processes alerts created from your searches:- Monitor for new alerts - Automatically detects alerts from completed searches or polls provider
- Gather intelligence - Enriches alerts with registry data (company records, officer information)
- Analyze matches - Uses AI to compare subject profiles against hit profiles with evidence-based reasoning
- Post resolutions - Automatically posts FALSE_POSITIVE or TRUE_POSITIVE determinations to your provider
- Track sync status - Updates alert remediation_status to reflect sync progress
Remediation Workflow
The remediation status tracks the alert through its lifecycle:- UNREMEDIATED - Alert created, awaiting AI analysis
- PENDING_SYNC - AI has made resolutions, waiting to sync to provider
- REMEDIATED_SYNCED - Resolutions successfully synced to provider
- REMEDIATED_UNSYNCED - Resolutions made but couldn’t sync (manual intervention may be needed)
How It Works
1. Alert Detection
The system continuously monitors your screening provider for new alerts. When a new case appears:- Alert is automatically ingested with full profile data
- All hits (potential matches) are retrieved
- Related articles and media are captured
2. AI Analysis & Enrichment
The AI engine enriches each alert with additional intelligence and performs deep analysis:- Gathers supporting data from public registries and databases
- Compares the subject against each potential match using multiple data points
- Identifies supporting and contradicting evidence
- Generates a confidence score and human-readable explanation
- Determines whether the match is a FALSE_POSITIVE or TRUE_POSITIVE
3. Resolution Posting
For each hit, the system:- Posts a comment and a resolution (FALSE_POSITIVE or TRUE_POSITIVE) to provider if so configured
- Includes summary explaining the determination
- Preserves any existing manual comments from your team
- Adds “Diligent AI: ” prefix for tracking
How the AI Makes Decisions
The AI evaluates each potential match by comparing multiple data points:- Identity information: Names, dates of birth, incorporation dates
- Location data: Addresses, countries of residence or operation
- Identifiers: Company registration numbers, tax IDs, passport numbers
- Contextual data: Industry codes, related entities, historical records
Example Resolution
Subject: Sarah Mitchell, DOB: 1988-03-15, UK Hit: Sarah MITCHELL, INACTIVE PEP, spouse of political figure (born 1965) AI Determination: FALSE_POSITIVE (Score: 0/100) Reasoning: “Temporal impossibility: Sarah Mitchell was born in 1988, but the hit profile indicates marriage to a political figure in 1992—when the subject would have been 4 years old. Additional evidence from public records shows the hit individual must have been born before 1965 to marry in 1992. These are different individuals with the same name.” Evidence:- ✅ MATCH (STRONG): Exact name match - “Sarah Mitchell”
- ✅ MATCH (MODERATE): Nationality - United Kingdom
- ❌ MISMATCH (STRONG): Temporal impossibility - marriage date incompatible with DOB
- ❌ MISMATCH (STRONG): Age discrepancy - 20+ year difference inferred
Provider Integration
The remediation agent works seamlessly with:WorldCheck (LSEG World-Check One)
- Monitors cases via date-based search
- Posts resolutions using WorldCheck resolution toolkit
- Maps AI determinations to configured risk/reason taxonomy
- Preserves existing remarks when updating
ComplyAdvantage CSOM
- Monitors searches via API pagination
- Posts resolutions as entity comments
- Prefixes AI comments with “Diligent AI: ” for tracking
- Supports both comment-only and full status updates
LexisNexis Bridger
- Monitors records via predefined search queries
- Posts resolutions to case assignment system
- Updates case status and adds remarks
- Maintains assignment role/division configuration
Configuration
Contact us at [email protected] to enable AI remediation for your account. To get started, please provide the following information for your screening provider:WorldCheck (LSEG World-Check One)
- API Key: Your WorldCheck API key
- API Secret: Your WorldCheck API secret
- Account ID: Your WorldCheck account identifier
- Group ID: The group ID to use when creating new cases (determines screening scope and available fields)
- Risk Label: Default risk category for resolutions (e.g.,
"LOW","MEDIUM","HIGH","UNKNOWN") - Reason Label: Default reason category for resolutions (e.g.,
"No Match","Full Match","Partial Match")
ComplyAdvantage CSOM
- API Key: Your ComplyAdvantage API key
- Search Profile ID (optional): Default search profile to use for new searches
- Region: API region - EU (
api.eu.complyadvantage.com) or US (api.us.complyadvantage.com)
LexisNexis Bridger
- API Key: Your LexisNexis Bridger API key
- Username: Account username in format
client_id/user_id - Password: Account password
- Predefined Search Name: Name of your predefined search configuration
- Assignment Role (optional): Default role to assign cases to
- Assignment Division (optional): Default division to assign cases to
AI Behavior Configuration
The remediation agent’s decision-making can be customized based on your risk appetite and compliance requirements. Configuration options include: Hit Category Rules- Define different behavior for different alert types (SANCTION, PEP, WATCHLIST, MEDIA)
- For example: conservative rules for sanctions (comment-only analysis) vs. permissive rules for adverse media (auto-resolve clear false positives)
- Set minimum evidence strength required for auto-resolution (weak, moderate, or strong)
- Configure separate thresholds for name evidence vs. additional evidence (DOB, location, identifiers, etc.)
- Use multiple rule segments per category for nuanced decision-making
- Auto-resolve: AI posts definitive resolution status to provider (e.g., mark as FALSE_POSITIVE)
- Comment-only: AI adds analysis as a comment without changing hit status (human reviews and decides)
- Configure different actions based on evidence confidence
- Conservative (sanctions): Comment-only for all cases, even with strong contradicting evidence
- Balanced (PEP): Auto-resolve slam-dunk false positives, comment on edge cases
- Permissive (media): Auto-resolve both clear false positives and clear true positives