Ron Norris

We hear about AI every day, but the truth is, not all AI is the same. What if you could have AI that learns directly from your subject matter experts, combining their knowledge with traditional data to understand how your business really works? Imagine having an AI capability that does the heavy lifting in near real time, pinpoints what is wrong, weighs each possible response, and shows you the action that delivers the strongest outcome with reduced risk and higher profitability.

That is the power of Causal AI.

Generative AI, Traditional Data AI, AI Workflows, Causal AI, and Agentic AI each solve different classes of problems. In energy, where complexity is high and decisions carry enormous risk, choosing the wrong type of AI can waste investment and weaken trust. Choosing the right combination, anchored by Causal AI, can transform operations, safety, decision-making, and other complex challenges through cause and effect.

• Understand the five main types of AI (Generative, Data, Workflows, Causal, Agentic) and how each one contributes to an integrated Agentic AI system 
• See how Causal AI combines expert knowledge with data to reason through cause and effect in complex operations 
• Learn how AI types work together, with Causal AI serving as the reasoning layer for real-time decision support 
• Explore practical use cases in energy that reduce risk, accelerate decisions, and increase profitability 

Most manufacturers are not short on data. They are short on data they can trust. Signals are noisy. Context is missing. Processes drift. Correlation models can amplify that mess and produce recommendations nobody will bet a line on.

This session shows how Causal AI separates true drivers from noise, fuses imperfect sources, and turns findings into interventions teams can defend and execute. The emphasis is practical. What question to ask first. What to do when the data is incomplete. How to use engineering knowledge as a guardrail. How to measure impact so the plant actually adopts it.

What we will cover

• The messy-data reality. Missing tags, sensor drift, recipe changes, shift effects, and why many ML models fail in production.
• From symptom to driver. Translating a quality or reliability problem into a causal question and a testable model.
• Making imperfect data usable. Data triage, stitching historian, MES, LIMS, maintenance, and quality records. Causal-guided pruning that improves signal without throwing away meaning.
• Expert knowledge as a control, not an opinion. Constraining discovery with process physics and engineering rules to avoid false confidence.
• Turning insight into action. Counterfactuals, safe-to-try interventions, and how to quantify impact on scrap, yield, downtime, and warranty.

Takeaways for attendees

A repeatable workflow to go from messy data to a decision you can defend. A simple checklist for data readiness and triage. A template for framing causal questions and validating results with operators and engineers. Clear guidance on measuring results and closing the loop.