AI-03 / INTELLIGENCE AI

Plintir Augur

Use AI to interrogate authorized intelligence, explain material change, identify gaps and prepare the next workflow—while analysts retain evidence and authority.

Core intelligence outcome
AI Decision Intelligence

Augur is designed to accelerate intelligence work without turning machine output into silent authority. Analysts can inspect the evidence, challenge the conclusion and decide whether a proposed action should move forward.

RETRIEVALCALLER-AUTHORIZED
OUTPUTCITED + CONFIDENCE
ACTIONHUMAN APPROVAL
Product experience

Intelligence in context, with the next decision already in view.

PLINTIR / AUGURLIVE
AI DECISION INTELLIGENCESTATUS: OPERATIONALUTC
RETRIEVALCALLER-AUTHORIZEDLIVE
OUTPUTCITED + CONFIDENCEVISIBLE
ACTIONHUMAN APPROVALCONTROLLED
CONFIDENCE0.86ASSESSED
Intelligence request
Summarize high-confidence changes in the priority entity network and identify unresolved intelligence gaps.
RPT-1209
ENTITY-00921
IMG-7714
COLLECT-042
Next intelligence action

Suggested collection request

GapLOCATION CONFIRMATION
Evidence6 SOURCES
RetrievalCALLER-SCOPED
PolicyREVIEW REQUIRED
Augur prepares the request. It cannot submit one. Execution is checked at the moment of action, never inherited from earlier reasoning.
Detailed capabilities

What Augur does.

CAP-01

Authorized intelligence retrieval

Apply the caller’s intelligence access before information reaches the model so answers are based only on evidence the user is allowed to use.

Access-aware retrievalNeed-to-knowSource scopeNo hidden post-filter
CAP-02

Cited, confidence-bearing answers

Attach evidence to material claims and visibly distinguish what comes from source reporting from what the model inferred or summarized.

Inline citationsConfidenceSource vs inferenceContradictions
CAP-03

Human-authorized intelligence actions

Allow AI to prepare collection requests, tasks, briefings and updates while keeping execution under named human authority and policy checks.

Propose actionsHuman reviewNamed authorityDecision record
CAP-04

Fit the model to the mission

Use different approved models for different intelligence tasks and handling contexts, with evaluation focused on accuracy, retrieval quality and safe behavior.

Model choiceEvaluationTask fitSafety testing
Interface detail

Model choice follows handling, not benchmark position.

PLINTIR / COGNITION · MODEL ASSURANCELIVE
ModelHostingAuthorized toTask accuracyRefusal qualityStatus
private-largeOn-premisesRESTRICTED
0.94
0.97
LIVE
private-compactForward nodeCONTROLLED
0.88
0.95
LIVE
vision-imageryOn-premisesCONTROLLED
0.91
0.93
LIVE
transcribe-multiSovereignOFFICIAL
0.86
0.90
LIVE
external-serviceCommercialPUBLIC ONLY
0.92
0.71
BLOCKED

The commercial endpoint scores well and stays blocked. Routing is decided by the handling level of the data involved, and evaluation continues after deployment rather than stopping at selection.

AI proposes. People decide.

Augur is designed to accelerate intelligence work without turning machine output into silent authority. Analysts can inspect the evidence, challenge the conclusion and decide whether a proposed action should move forward.

Intelligence flow

From information to Actionable Intelligence Data.

STEP 01

Ask

The analyst asks a mission question in natural language.

STEP 02

Retrieve

Authorized evidence is selected from the intelligence context.

STEP 03

Explain

Augur produces a cited answer with uncertainty and contradictions visible.

STEP 04

Prepare

The system can prepare a next step for authorized human review.

Mission use cases

Designed around intelligence outcomes.

UC-01

Intelligence Copilot

Summarize large intelligence collections with evidence attached.

UC-02

Watch-Floor Change Brief

Explain what materially changed since the prior shift and what remains unresolved.

UC-03

Commander Intelligence Update

Prepare a concise brief with cited evidence, gaps, assumptions and recommended follow-up questions.

Product briefing

See Augur against a real intelligence workflow.

Bring a recurring intelligence question that currently requires multiple analysts to search, reconcile and brief across disconnected sources.