The Intelligence Center

A three-surface system for detecting, validating, and understanding baseline deviation — from raw signal triage through to causal impact analysis.

Observations (Events)

Operational realities detected during execution. They represent state changes—a delayed delivery, a safety incident, or a site log entry. They require triage, linking to specific activities, and Carry dynamic exposure.

Requires Triage

Project Constraints (Rules)

Ambient rules that govern the project. They represent static truth—liquidated damages, weather entitlements, or notice windows. They don't change state, but they define the mathematical parameters of impact.

Forensic Truth

The Three Surfaces

Each surface has a single, clear responsibility. Data flows forward — Feed → Activity Panel → Executive Layer — but never backward.

1. Intelligence Feed

A pure triage inbox. Shows only detected signals — observations the engine has raised but you haven't reviewed yet. Accepted signals leave this surface permanently.

Verify → moves signal to causal record

2. Activity Detail Panel

The causal truth surface. Two tabs: Overview (CPM metrics, float, logic trace, AI strategy) and Contributing Signals (accepted signals linked to this activity with locked exposure and delay values).

Immutable once accepted

3. Executive Layer

Aggregated KPIs and heatmaps for leadership. Three heatmap modes: Baseline Variance, Financial Exposure, and Float Erosion. Driven by accepted model artifacts — not raw signals.

Read-only, always current

Signal Lifecycle

Every signal from the Oracle follows a strict one-way state machine. Once accepted, its values are frozen as model artifacts and can never be retroactively edited.

DETECTED

Engine raised observation

LINKED

Mapped to activity

REVIEWED

Under human review

ACCEPTED

Locked model artifact

DISMISSED

Removed from model

Rejected signals never appear in activity context or the Deviation View impact drivers. They are removed from the causal model entirely, keeping causality clean.

Key Capabilities

Human-in-the-Loop Triage

Every signal requires a human decision. The AI proposes; you validate. Verified signals carry a timestamp and user attribution.

CPM Impact Preview

Before accepting a signal, see its projected CPM impact — predecessor logic, critical path effect, and estimated schedule slip — anchored to the current Data Date.

Locked Exposure Values

When you accept a signal, its financial exposure and delay values are frozen (lockedExposureValue, lockedDelayValue). These become audit-grade model artifacts.

Baseline Drift Banner

A persistent authority strip shows Data Date, active Baseline, and project-wide drift. Clicking the banner opens the baseline comparison layer across the schedule.