Intelligence Layer

Trunnion AI

The native AI engine embedded into every module, dashboard, and workflow within Cortrova.

Developed by Trunnion AI, a subsidiary of Viceroy NM. Not a bolt-on — the nervous system of the entire platform.

0
AI Agents
3
Agent Tiers
6
Analysis Engines
0
Domain Analyzers
0
Departments
Architecture

Three-tier agent hierarchy.

Bottom-up intelligence pipeline — each tier feeds the next, creating a comprehensive operational intelligence network.

51

Tier 1 — Department Agents

3 agents × 17 departments

Analyst detects anomalies on dashboard load (~8K token context). Advisor generates recommendations for individual records (~4K tokens). Reporter produces KPI summaries every 30 minutes (~6K tokens).

7

Tier 2 — Division Aggregators

Cross-department synthesis every 30 min

Seven divisions synthesize insights across related departments: Operations & Production, Safety/Regulatory/Environmental, Finance & Procurement, Supply Chain & Logistics, People & Facilities, Commercial & Customer, Technology & Innovation. ~10K token context.

6

Tier 3 — Executive Agents

5 specialists + 1 synthesizer

Strategy, Operations, People & Culture, Risk & Compliance, Market Intelligence — each with 12K token context. The Synthesizer combines all into a unified C-suite briefing (16K tokens, 4K output).

Analysis Engines

Six engines. One intelligence fabric.

Running every 6 hours and on-demand, querying 90 days of data across all 17 departments.

📊

Predictive Alert

Linear regression with R² scoring, 7/30-period SMA deviation with Z-score anomaly detection (±1.5σ), threshold proximity alerting within 10%. Severity + confidence scoring on every alert.

🔍

Root Cause Analysis

Cross-department correlation detection. When quality drops, traces to overdue maintenance, supplier changes, staffing shifts — connections invisible in siloed systems.

Prescriptive Recommendations

60+ actionable templates with real metric values, effort/cost estimates, historical effectiveness tracking, and projected downstream KPI impact.

🔮

Scenario Modeling

50+ pre-seeded causal relationships with impact coefficients, confidence scores, and lag periods. Model interventions before committing resources.

Resource Optimization

Multi-issue prioritization scoring urgency, impact, availability, and cross-department dependencies. Prioritized action plans with assigned owners.

📈

SPC Anomaly Detection

Continuous Cpk/Cp monitoring, 1σ drift detection, 3σ out-of-control flagging — before defective product ships. Runs 24/7 with no manual chart review.

Orchestration

Automated intelligence on a schedule.

Every 15 min
Department health scores computed for all 17 departments (0–100 composite)
Every 30 min
Reporter agents + Division aggregators generate cross-department insights
Every 6 hours
Full analysis pipeline — all 10 analyzers + cross-domain correlation + predictive alerts
On demand
Executive briefings, Narrator conversations, mock audits, ad-hoc analysis
Security

Multi-layer AI governance.

Every request: Team Guard → Kill Switches → Rate Limiting → Budget Check → Scope Validation → LLM Call → Audit Log.

Kill Switches

Three levels — global, division, team. Instant shutdown with zero caching delay.

Scope Registry

Declarative data boundaries. Safety agents cannot query Finance models.

Rate Limiting

Per-user, per-team RPM via Redis. Default 60 RPM, configurable.

Token Budgets

Daily budget per team. Warning at 80%. Hard stop at 100%.

API Key Mgmt

Per-team encrypted keys with rotation and model allowlists.

Audit Trail

Every agent action logged. Non-blocking. Fully queryable.

Get Started

See Trunnion AI in action.

Watch 65 agents analyze your operational data in real-time.

FAQ

Trunnion AI Questions

Is Trunnion AI included or an add-on?

Fully included — all 65 agents, 6 engines, and signature capabilities at no additional cost.

Can I disable AI for specific departments?

Yes. Three kill switch levels — global, division, team — with instant effect and zero caching delay.

What LLM models does it use?

Anthropic Claude and OpenAI GPT via provider abstraction. Different departments can use different models simultaneously based on task requirements and cost optimization.