How It Works

From intake to IAPs — fast, grounded, and human-in-the-loop

TEW state machine + fusion workbench + decision support. Automation handles the flood; people make the calls that matter.

Automation~75%Intake → Draft products
Core StatesAlert → Warning → Response → OutbreakTEW workflow engine
On-Prem AILocal LLMRAG + Guardrails
OutputsAdvisories • IAPs • SITREPsIC/UC-ready

The end-to-end flow

Eight stages. Automation accelerates; humans approve.

1 • Intake & Triage

Consolidate chatter, calls, texts, emails

STT/NLU, dedup & clustering, risk scoring. Analyst promotes credible clusters to incidents.

2 • TEW State Machine

Alert → Warning → Response → Outbreak

LLM proposes transitions, EEI checklists, notifications. IC/UC confirms state change.

3 • Fusion Workbench

IPO + COAs + Collection Plan

Maps, RIFs, playbooks, intel streams. LLM drafts intel estimate & COAs; analysts refine and task cells.

4 • Net Assessment

Decision support

Scope, magnitude, consequences, second-order effects, RAMs. IC/UC approves & publishes.

5 • Products

Advisories • IAPs • SITREPs

Auto-drafted with citations, redacted by role. Final sign-off before release.

6 • Notifications

Calls • SMS • Email

Matrix-driven outreach with delivery logging. Sensitive public/media comms double-checked.

7 • C-UAS / Port Ops

Sensor fusion + SOPs

Track classify, operator-hunt flow, authority lanes. Interdictions remain a human authorization.

8 • After-Action

Learn & codify

AAR timeline is auto-built; validated lessons feed back into playbooks and RIFs.

Automation & human lockpoints

Where the platform runs hands-free — and where humans stay in charge.

Phase Automation What’s automated Human lockpoints
Intake & triage 85–90% OSINT scrape, STT/NLU on calls, dedup/clustering, risk scoring Promote to Incident
State machine 70–80% Proposed transitions, EEI checklists, notify matrix Approve Alert/Warning/Response
Fusion workbench 60–70% Intel estimate draft, COAs, collection plan Edit/assign tasks
Net assessment 70–80% Scope/magnitude, consequences, RAMs Publish decision support
Products 80–90% Draft Advisories, IAPs, SITREPs; redaction by role Approve final docs
Notifications 80% Calls/SMS/email via matrix; delivery logs Approve public/media comms
C-UAS / Port ops 70% Sensor fusion, track class, AAR build Authorize interdictions
After-action 60–70% Timeline, findings, suggested playbook updates Validate lessons learned
Lockpoint #1

State transitions

Alert/Warning/Response approvals remain IC/UC decisions (“Confirm & notify”).

Lockpoint #2

Public/media messaging

Human review for tone, clarity, and authority; auto redaction for role-scoped versions.

Lockpoint #3

Use of force / C-UAS actions

Automations advise; command authorizes based on legal/authority lanes.

Local AI, grounded in your doctrine

On a single AI Command Center or distributed nodes.

Models

120B for heavy lifts, 7–13B for speed

Router chooses model by task. All prompts constrained by role and policy.

Grounding

RAG over Playbooks, RIFs, Mission Folders

Outputs cite sources; no free-wheeling text. Everything is auditable.

Privacy

RBAC/ABAC + redaction

PII/CJI filters pre/post generation; chain-of-custody for every product.

See it in action

Download the starter kit or request a guided demo scenario.

Download Request Access