Sentinel shows what happens between an AI agent being able to act and an enterprise deciding that it may act. Inspect the proposal, authorization gate, human authority, execution permit, protected action, and evidence trail in motion.
View CI test resultsSynthetic data onlyInteractive 3D Decision X-RayProtected execution permitEvidence ledger
A synthetic browser visualization, separate from the executable Python reference engine. This page does not prove production enforcement. Review release gates and limitations.
The agent may propose the state change, but capability does not supply authority. No execution permit exists until the required human approval is present.
Decision trace
Workflow intelligence
17 human steps become a governed human–agent workflow.
The example is fictional. Sentinel does not ask “where can we add AI?” first. It models the work, identifies where bounded automation is useful, and preserves human authority where the consequence justifies it.
AgentAI assistHuman authority
Baseline
17 human touches. Sequential processing. Partial decision evidence.
Governance should prove whether the redesign helped.
These are synthetic design targets, not production or client claims. The purpose is to make workflow transformation measurable before AI scale is declared successful.
-80.21%cycle time
-64%cost units
-73.33%quality defects
-64.71%human touches
+55 ptsevidence completeness
Evidence engineering
The decision leaves a reconstructable trail.
The Python reference engine can export effective policy and delegated authority, the control trace, and execution events correlated by decision ID. A permit records authorization; it does not prove successful execution. The browser animation illustrates these concepts.