01 HUMAN DECIDES
Advise
The ranked root cause and recommendation are presented to an operator. A human decides and acts. Best for early rollout and high-consequence calls.
TRUST & ENTERPRISE READINESS
Governed by design
A ranked causal result never goes straight to action. It passes explicit gates before it reaches an operator, or acts within bounded limits — governance as the enabler of autonomy, not a brake.
Confidence-scored. Bounded. Human in control.
← Back to Real-Time Causal AITHE GOVERNED PATH
From ranked result to governed action
Every result passes explicit gates before it advises an operator or acts within bounds. Nothing skips the path.
INPUT Ranked causal output
GATE 01 Reason over objectives
GATE 02 Deterministic policy
GATE 03 Independent safety authority
GATE 04 Confidence gate
Advise operator Act within bounds
Aligned with Functional safetyAlarm managementCybersecurityManagement of change
AUDITABLE BY DEFAULT
Every decision leaves a trace
Not a log buried in a system — a structured record a reviewer, an auditor, or a safety authority can follow end to end.
01 WHAT WAS SEEN The live anomaly and the plant state at the moment of the deviation.
02 WHY IT MATTERED The ranked causal factors and their confidence, drawn from the validated model.
03 WHAT GOVERNED IT The objectives, policy limits, and safety checks the result passed through.
04 WHAT WAS DECIDED The recommendation or action, who approved it, and the outcome.
HUMAN IN CONTROL
You set how far it can act
The same governed path routes to advice, approval, or bounded autonomy. You choose the level per decision — and move it as trust is earned.
02 HUMAN APPROVES
Approve
The system prepares the corrective action and its evidence. A human approves execution. Best for mature advisory workflows.
03 ACTS WITHIN LIMITS
Bounded autonomy
Within governed policy limits the action executes, and exceptions escalate. Best for well-understood decisions with mature controls.
RUNS WHERE DECISIONS HAPPEN
At the edge, or in the cloud, always governed
The same DataStreams engine and the same governance, whether it runs at the control-system edge or across the fleet in the cloud.
EDGE
At the control-system layer
- Local causal models run inside the plant
- Ranked root cause in minutes, on live process data
- For real-time operator response
- Air-gappable, with an open-weight model on local hardware
CLOUD
Across the fleet
- Fleet-wide causal analysis in the cloud
- Find root cause before failure for reliability & maintenance
- For better planning across sites
- Integrates with maintenance, ERP & lab systems
RUNS ON LIVE, MULTI-SYSTEM DATA
Time-series Structured assets Unstructured procedures & hazard studies A semantic ontology
THE FOUNDATION FOR SAFE AUTONOMY
Governance is what makes autonomy safe.
Confidence-scored, bounded, and auditable — with a human setting the limits. See how the model is built, or talk through your governance requirements.
