Skip to main content

Understand Tab

Strategic Environment Analysis — JP 5-0, Step 1

Purpose​

The Understand tab is the entry point for every problem set in BASTION. It ingests raw documents, extracts structured intelligence, and builds the shared understanding that drives all downstream planning. This corresponds to JP 5-0 Step 1: Planning Initiation / Understanding the Operational Environment.

Staff use this tab to upload source material, review AI-extracted entities and relationships, validate strategic objectives, and monitor the evolving intelligence picture through the RAFT knowledge graph.

Components​

Document Upload​

  • Accepts PDF and DOCX files.
  • On upload, AI agents automatically extract entities, relationships, and key themes.
  • Documents are stored with full provenance metadata (uploader, timestamp, classification).

Scenario Package Upload​

  • Upload a bundled scenario package (multiple documents representing a single exercise or contingency).
  • AI tag inference automatically categorizes documents by domain (political, military, economic, etc.) and phase.
  • Packages seed the knowledge graph in bulk, accelerating initial understanding.

Autonomous Document Intelligence Team​

The Understand tab runs a 10-specialist AI agent team that processes uploaded documents in parallel (Reference: Phase 40):

AgentSpecialty
Strategic SynthesisCross-document theme extraction and overarching narrative
Entity ResolutionDeduplication and merging of entity references across sources
OSINT MonitorContinuous monitoring for new relevant information
Validity AssessmentConfidence scoring and unsupported-claim flagging
Conflict DetectionContradiction identification between sources
RAFT ExtractionKnowledge graph construction from documents
RAFT ReasoningGraph traversal for non-obvious inference
PMESII AnalystPMESII-PT dimension population and gap analysis
Source ReliabilityNATO source reliability rating application
Objective ExtractorStated and implied objective identification with traceability

Agents operate in parallel on uploaded documents; their outputs are aggregated before staff review, dramatically reducing initial intelligence fusion time.

ExtractionTheater Live Pipeline Visualization​

  • Real-time view of the document intelligence pipeline as it runs.
  • Shows which agents are active, what document sections they are processing, and what entities have been extracted so far.
  • Allows staff to intervene early if extraction is going in the wrong direction.
  • Completion state displays a structured summary of all extracted elements before any enter the knowledge graph.

Brain Visualization (Knowledge Graph Canvas)​

  • Adaptive force-directed neural canvas rendering the full knowledge graph for the current problem set (Reference: Phase 41).
  • Nodes are color-coded by type: actors (blue), locations (green), events (orange), objectives (yellow), documents (grey).
  • Edges show relationship type and directional strength.
  • Brain Timeline allows temporal filtering — view the graph at any point in its extraction history.
  • Subspace concept manages graph scale: large problem sets can be divided into focused subgraphs for specific staff sections.
  • JSON-LD semantic triples underpin the graph, enabling ontology alignment and interoperability with external intelligence systems.

Scoping Interview​

  • Structured AI-facilitated interview before document upload to scope the problem set.
  • Captures: mission type, geographic focus, doctrinal framework, classification level.
  • Interview outputs pre-configure extraction filters and agent team parameters.
  • Reduces false extractions and focuses the document intelligence pipeline on relevant entities and relationships.

RAFT Graph Visualization​

  • Interactive force-directed graph of extracted entities and relationships.
  • Nodes represent actors, objectives, capabilities, and constraints.
  • Edges represent relationships such as supports, opposes, enables, and constrains.
  • Filter by entity type, domain, or confidence score.

NATO Source Reliability Ratings​

  • Each intelligence input is rated on the NATO Admiralty Code:
    • Source Reliability: A (Completely Reliable) through F (Reliability Cannot Be Judged)
    • Information Accuracy: 1 (Confirmed) through 6 (Truth Cannot Be Judged)
  • Ratings are applied by the Source Reliability agent and can be overridden by J2.
  • Confidence scores in the knowledge graph incorporate both the source reliability and information accuracy ratings.
  • Ratings are visible on all extracted entities to support staff judgment.

PMESII-PT Analysis​

  • Structured breakdown across all eight PMESII-PT dimensions: Political, Military, Economic, Social, Information, Infrastructure, Physical Environment, and Time.
  • AI agents populate initial assessments; staff refine and validate.
  • Feeds directly into the Design tab's problem framing canvas.

Strategic Objective Extraction​

  • AI identifies stated and implied objectives from uploaded documents.
  • Objectives are linked to source passages for traceability.
  • Staff approve, reject, or modify extracted objectives before they flow downstream.

Validity Dashboard​

  • Displays confidence scores and source coverage for each extracted element.
  • Flags gaps in intelligence coverage and contradictions between sources.
  • Tracks staleness — how recently each assessment was updated.

AI Agents​

The Understand tab deploys a 10-specialist autonomous document intelligence team (see Components section above). All agents operate under human oversight through the ExtractionTheater pipeline visualization. Extracted outputs require staff review before they become authoritative inputs to downstream tabs.

Key agent functions:

AgentFunction
Strategic SynthesisSynthesizes across multiple documents to identify overarching themes and connections.
Entity ResolutionDeduplicates and merges entity references across documents (e.g., recognizing "PRC" and "China" as the same actor).
OSINT MonitorWatches for updates and new information relevant to the current problem set.
Validity AssessmentScores confidence levels and flags unsupported claims.
Conflict DetectionIdentifies contradictions between sources or between extracted elements.
RAFT ExtractionParses documents to build the Retrieval-Augmented Fine-Tuning knowledge graph.
RAFT ReasoningTraverses the knowledge graph to surface non-obvious connections and inferences.
Source ReliabilityApplies NATO Admiralty Code ratings to all intelligence inputs.

Role Access​

  • All staff roles have read access to the Understand tab.
  • J2 Intelligence has primary responsibility for validating extracted intelligence and maintaining the knowledge graph.
  • Commander and J5 Plans review strategic objectives before handoff to Design.

Data Flow​

Documents / Scenario Packages
        |
        v
  AI Extraction & Analysis
        |
        v
  +--------------------------+
  | Extracted Objectives     |
  | RAFT Knowledge Graph     |
  | PMESII-PT Assessments    |
  | Intelligence Estimates   |
  +--------------------------+
        |
        v
  Design Tab (Problem Framing, CoG Analysis)

Inputs​

  • Raw documents (PDF, DOCX)
  • Scenario packages
  • External intelligence feeds (via OSINT monitor)

Outputs​

  • Validated strategic objectives
  • RAFT knowledge graph
  • PMESII-PT structured assessments
  • Intelligence gaps and confidence scores

These outputs feed the Design tab as the foundation for operational design.

Doctrinal Reference​

  • JP 5-0, Chapter II: Planning Functions — Understanding the Operational Environment
  • JP 2-0, Joint Intelligence
  • ADP 5-0, The Operations Process — Understanding

Part of the BASTION Capability Tabs documentation.