II
IndustrialIQ
Industrial Metals Operations Intelligence · OT/IT · MES/ERP · Historian · Unified Namespace
AI learning from 2,847 operational outcomes
ENV: PROD · ROLE: OPERATIONS LEAD
Dashboard
TrendIQ 10
DecisionIQ 8
ActionIQ 8
EffectivenessIQ 27
StaffingIQ 8
Architecture
Industrial Metals Operations
Continuous operational intelligence across plants, melt shops, casters, and rolling mills
◊ Knowledge graph: 62,481 linked operational artifacts
Operational Stability
92.4%
↑ 3.8pts vs prior quarter
Unified Namespace Coverage
81%
↑ from 68% (Q1)
Signal-to-Action Closure
78%
↑ 11pts vs Q1
Time-to-RCA −36%
−36%
From 48hr → 31hr median
Open Operational Risks
6
3 critical, 3 high
Continuous Intelligence Loop
Signal → Decision → Execution → Outcome → Organizational Learning
● Active across all modules
TrendIQ
10 active signals
→
DecisionIQ
8 initiatives
→
ActionIQ
7 in execution
→
EffectivenessIQ
27 outcomes learned
Time-to-RCA — 12-week trend
Median hours from signal detection to root-cause identification, across all plants
Engineer Time Recaptured
Hours/week saved from manual correlation, dashboarding, and triage
Cross-System Signal Coverage
Share of operational signals carrying full cross-system context (OT + IT + MES + ERP)
Signal Mix — Last 30 Days
Distribution of signals by operational domain
AI Confidence Distribution
Confidence-tier breakdown across active recommendations
Top Operational Risks — AI-prioritized
Highest-impact items pulled from TrendIQ
AI Operational Copilot
Live workflow optimizations the system is recommending
TrendIQ — Operational Signal Detection
Real-time signal detection across PLC, SCADA, historian, MES, ERP, and vision systems — AI surfaces cross-system patterns invisible to any single dashboard
● Connected to 14 industrial source systems
Active Signals
10
↑ 3 this week
Pattern Clusters
5
2 emerging
Cross-Domain Correlations
6
3 high-strength
RCAs in Progress
4
Avg conf 86%
AI Recommendations
5
Pending review
Signals 12
Patterns 5
Correlations 7
Root Cause Analysis 4
Recommendations 9
DecisionIQ — Initiative Management
Active operational initiatives created from accepted signals, with AI-suggested follow-on tasks
8 active initiatives 6 AI suggestions pending
AI Continuously Recommending
The system is learning from 2,847 historical operational outcomes to improve every recommendation below
Initiative owners
94% acceptance
Task owners
89% acceptance
Closure paths
82% accepted
Follow-on tasks
71% accepted
ActionIQ — Execution Coordination
Two-tier Kanban: initiatives at the top, task drill-down on click
Initiatives Kanban
AI Operational Copilot
Live workflow optimizations across the board
Automation Opportunities
Workflows AI can orchestrate end-to-end (with operations lead signoff)
EffectivenessIQ — Organizational Learning
What worked, what failed, and how the system is getting better at operational recommendations
● Continuous learning active
Outcomes Captured
2,847
↑ 184 this month
SME Corrections
146
→ model refinement
Initiative Success Rate
78%
↑ from 61% (Q1)
AI Recommendation Accuracy
87%
↑ 12pts vs baseline
Failed Closures Studied
19
Root-caused
Recent Learning Events
Outcomes from closed initiatives feeding back into the recommendation model
Improvement Over Time
Recommendation acceptance rate by quarter
What's Working (Reinforced)
Patterns the AI is doubling down on
What's Not Working (Down-weighted)
Patterns the AI is moving away from
Add Custom Task
AI has pre-filled this based on the initiative's closure path
AI Pre-fill — Suggestions below are editable. Owner defaults to AI's recommendation.
. Fill in placeholder tokens before delivering. PLACEHOLDER TOKENS: industrial operations — domain label e.g. "AV program" / "manufacturing" / "DevOps" 8 — number of teams being monitored e.g. "12" 153 — total engineers across all teams e.g. "186" KPI strip values are rendered from siqTeams data — no tokens needed. Sub-tab content is fully JS-rendered from data arrays. ============================================================ -->
StaffingIQ — Team Capacity & Workforce Intelligence
AI continuously monitoring team load, skill gaps, attrition risk, and hiring requirements across industrial operations domains
● Analyzing 8 teams · 153 engineers
Teams Over Capacity
7
↑ 3 vs last quarter
Avg Team Load Index
118%
↑ 14pts — above threshold
Critical Skill Gaps
11
Across 5 domains
Open Headcount Needed
24
To restore healthy load
Attrition Risk — High
9
Engineers flagged
Team Overview
Capacity Signals
Demand Forecast
Recommended Actions
Team Load Heatmap — click any card to drill in
Load Index = (Active workload hours) ÷ (Available capacity hours) × 100. AI flags teams above 105%.
● >120% overloaded ● 110–120% stressed ● 100–110% at capacity ● <100% healthy
Load Index Trend — Last 6 Quarters
Average team load across all AV program domains
Headcount Gap by Domain
FTEs needed to restore load index below 100%
AI-detected capacity and staffing risk signals — ranked by program delivery impact
Workload Demand Forecast — Next 4 Quarters
AI projection based on program roadmap, ODD expansion, and historical velocity patterns
● AI projection
Hiring Priority Queue
Roles AI recommends sourcing first based on bottleneck impact
Skill Coverage Matrix
Current coverage vs. program-required proficiency by domain — red cells indicate critical gaps
11 critical gaps identified
AI-recommended actions to improve team capacity, reduce load index, and close skill gaps — ranked by delivery impact
Architecture — Unified Operational Intelligence Stack
The 5-layer architecture that turns fragmented OT/IT signals into closed-loop operational intelligence
● 14 industrial source systems integrated
From fragmented signals to operational intelligence
Industrial plants already generate enormous operational data — PLCs, SCADA, historians, MES, ERP, vision systems, robotics, quality systems. The challenge is no longer data collection: it is transforming fragmented signals into scalable operational intelligence. This architecture is the substrate the four IQ modules run on. Each layer below builds on the one above it, and the Continuous Learning Layer feeds outcomes back to every prior layer.
Layer 1
Industrial Signal Layer
Raw operational signals from the physical plant — sensors, PLCs, robotics, vision systems, instruments — collected through standard industrial protocols and brought into a unified namespace so signals carry plant, line, asset, and process context with them.
PLCs SCADA Historians (PI, Ignition) MQTT / OPC UA Vision systems Robotics Unified namespace
Foundation
Raw signal ingest
+ context tagging
Layer 2
Context & Semantic Layer
Operational models — asset hierarchies, process flow, product genealogy, ISA-95 / ISA-88 ontologies. This is where a "tag" becomes a meaningful signal tied to a specific furnace zone, caster strand, rolling stand, or cold-strip pass with full upstream/downstream context.
Asset hierarchy (ISA-95) Process genealogy Equipment ontologies Product / batch tracking MES / ERP join Master data alignment
Context
Operational models
+ semantic joins
Layer 3
Intelligence Layer
AI-driven trend detection, anomaly analysis, cross-domain correlation, RCA, and recommendation. This is where TrendIQ surfaces signals and DecisionIQ ranks initiatives. Models are trained on contextualized signals from Layer 2 — never on raw, contextless time-series alone.
TrendIQ — signal detection DecisionIQ — initiative ranking Anomaly clustering Cross-domain correlation RCA inference Recommendation engine
Intelligence
TrendIQ + DecisionIQ
run here
Layer 4
Action Layer
Workflow orchestration, escalation routing, work-order creation, and operator HMI integration. This is where ActionIQ coordinates execution — translating ranked initiatives into assigned tasks, automated work orders, and HMI prompts that surface AI guidance at the point of operator decision.
ActionIQ — Kanban orchestration CMMS work-order trigger Operator HMI prompts Escalation routing Cross-team coordination Closed-loop controls
Action
ActionIQ runs here
+ point-of-decision UI
Layer 5
Continuous Learning Layer
Outcome capture, SME corrections, and feedback loops. Every closed initiative — what worked, what failed, where senior operators corrected the model — feeds back into Layers 2 and 3. This is the differentiator: the platform gets demonstrably better over time, not just initially instrumented.
EffectivenessIQ Outcome scoring SME correction capture Model refinement Pattern promotion / demotion Institutional learning
Learning
EffectivenessIQ
feeds back to all layers
Source Systems Integrated
The platform pulls signals and context from across the OT/IT, MES/ERP, and engineering-tooling estates
14 systems connected
Operational Technology

OT Layer

  • PLCs (multi-vendor)
  • SCADA
  • Sensors & instruments
  • Vision systems
  • Robotics
  • Industrial networks
Connectivity & Middleware

Bridging Layer

  • Kepware
  • OPC UA
  • Ignition
  • MQTT / Sparkplug B
  • Unified namespace
Historians & Data Platforms

Time-Series Layer

  • OSIsoft PI
  • Ignition Historian
  • IBA systems
  • SQL Server
  • Lakehouse / data lake
Manufacturing & Enterprise

MES / ERP Layer

  • MES platforms
  • ERP (production + supply)
  • Quality systems
  • CMMS / EAM
  • Supply chain platforms
Architecture Principles
Non-negotiables that the 5-layer stack is built on — and that any new component or initiative must respect
Context-first ingestion
No signal enters the platform without its plant / line / asset / process context. Raw tags without context are not actionable, and they bias downstream ML models.
OT/IT signal handoffs are explicit
Every signal contract between the OT side (PLCs, historians) and the IT side (MES, ERP, AI services) is named, owned, and versioned — not implicit, not point-to-point.
Closed-loop, not open-ended
Every recommendation has a path to action and a captured outcome. The platform fails when initiatives close without their outcomes flowing back to Layer 5.
Multi-site portable from day one
Architectural patterns must work across plants from the start. Per-site customization is technical debt; semantic alignment is the asset.
Operator at the point of decision
AI recommendations land in the HMI the operator already uses — not in a separate dashboard. Operators see, accept, override, and feed back from one surface.
SME feedback is first-class
Senior operator and engineer corrections are not edge cases — they are the highest-value training signal the platform receives. They flow back to Layers 2 and 3 directly.