OverviewAI Analysis Engine

AI Analysis Engine

AI that speaks intelligence

How It Works

Step through the process to see how it works.

1

Data Gathering

The AI engine collects all relevant entity data — profiles, relationships, events, communications, and financial records.

AI Analysis Pipeline

Data Gathering

Entity profiles, relationships, events

Person records
Event history
Network links
Location data
Stage 1

Prompt Construction

Context assembly + instructions

System role definition
Entity context block
Analysis instructions
Output format spec
Stage 2

LLM Processing

Cloud API or local Ollama model

Token processing
Pattern recognition
Inference generation
Response streaming
Stage 3

JSON Parsing

Structured output extraction

Schema validation
Field extraction
Error handling
Type casting
Stage 4

Report Generation

Intelligence products

Executive summary
Detailed analysis
Threat indicators
Recommendations
Stage 5
Active
Complete
Pending
Step 1 of 5

Key Features

Core capabilities that power this module.

Automated Dossiers

Comprehensive intelligence profiles generated from all available data — criminal history, network analysis, threat assessment.

Natural Language Query

Ask questions in plain English — "Which suspects in District 5 changed communication patterns this month?"

Anomaly Detection

AI identifies unusual patterns across communications, financials, and movement data that human analysts might miss.

Cloud/Local Switching

One config change swaps between cloud API (demo) and local Ollama model (air-gapped) — same interface, same output.

Scenario: AI Dossier on Ahmed Khalil

A realistic walkthrough using fictional intelligence data.

A senior analyst requests an AI-generated dossier on Ahmed Khalil. The system ingests 142 data points from across the platform and produces a structured intelligence report in under 30 seconds.

1AI ingests: 3 FIRs, 14 communication records, 5 financial transactions, 8 location pings, and 23 relationship edges
2Dossier includes: Executive Summary, Key Associates (5 persons ranked by connection strength), Threat Assessment (HIGH, 63/100)
3Recommended Actions: increase surveillance on 2 associates, request CDR deep-dive on 3 flagged numbers
4Natural language query: "Which Cell-7 members resumed contact this month?" returns 4 results with dates and evidence
5Same query runs identically on cloud API (demo) and local Llama 3.1 model (offline deployment)
Threat AnalysisBack to OverviewPrediction Engine

Ready to Transform Your
Intelligence Operations?

Deploy CLERINT Fusion on your infrastructure — cloud, on-premise, or fully air-gapped. See it in action with your own data.

Schedule a DemoContact Sales