The Industrial AI Sovereign: Read the doctrineThe Industrial AI Summit, Houston: Read the highlights

The Industrial AI Summit, Houston, November 4, 2026

Highlights from The Industrial AI Summit in Houston, and the Definitive Guide to Industrial Agentic AI, coming next week.

A full room at The Industrial AI Summit in Houston, with a speaker presenting beside two screens
Microsoft Innovation Hub, Houston, November 4.

On November 4, operators, CIOs and Microsoft’s industrial leaders filled the Microsoft Innovation Hub in Houston for The Industrial AI Summit. Every speaker faced the same test: show it working on real plant data. The questions were sharp, the demo ran live, and the conversation ran past two o’clock.

IntuigenceAI is a proud partner of The Industrial AI Summit. Thank you to Microsoft for hosting us, and to every speaker and guest who made the day.


Moe Tanabian giving the opening keynote
Moe Tanabian opens the day.
01

Opening keynote: Six success elements in industrial AI

Plant data needs a knowledge graph, not a bigger model.

Moe Tanabian, Founder and CEO, IntuigenceAI

  • Onboarding no longer takes a year. Building a knowledge graph for a typical plant of about 500 assets used to take nine to eighteen months. With AI reading the P&IDs, PFDs, time series and tens of thousands of documents, a working graph now takes days.
  • Zero egress skips the longest approval cycle. All 17 components, the AI models included, are licensed into the customer’s own Azure tenant. Nothing leaves, so the three to six months of CIO, CISO and legal review that egress triggers never starts.
  • Frontier models fall off a modality cliff on plant data, and charge more for it. Models that score 80 to 90 percent on text answered fewer than 10 percent of connectivity questions across 20 P&IDs. Domain-trained smaller models answer at 10 to 100 times lower cost, and a graph answers in milliseconds where a model reasons for minutes.
“If you want to build a demo, build it with a large language model. If you want to give it to 50 engineers in the field, build a knowledge graph.”Moe Tanabian
Bill Braun presenting lessons from Chevron's digital transformation
Bill Braun on three decades of digital transformation at Chevron.
02

Keynote: Key steps from Chevron’s digital transformation

Nobody gets to sit this out. Make fewer, bigger bets.

Bill Braun, Former CIO, Chevron

  • The map changes every six months, so leave markers for each other. Share what is not competitive, and pick up what others have learned. Industry groups that learn together move faster than any company alone.
  • Choose your stack, your model strategy and your partners deliberately. The model you use for an HR application is not the one you train on time series, and neither is the one for geology. Different paths need different partners.
  • Measure value in a few big bets, not in hours saved. Five hours a week per employee may never show up on the bottom line. The best results, some returning 20 times, come from defined processes with concrete measures, and leading companies make bigger, longer bets there.
“Nobody in the industrial space gets to not go on this journey.”Bill Braun
Christoph Berlin of Microsoft presenting on the fifth industrial revolution
Christoph Berlin on industrial AI with Microsoft Fabric.
03

Keynote: Accelerating AI with Microsoft Fabric for industrial use cases

Protect the knowledge. The models will come and go.

Christoph Berlin, Vice President, Engineering, Microsoft

  • It is a mindset, not a technology stack. Every previous revolution changed how work is done, and this one is no different. At a Stuttgart automaker, root-cause investigations that took nine months, mostly spent finding who held the data, now point teams to the evidence in minutes.
  • Let the data stay where it is. Put one layer of governance over it. Fifty databases are not the problem; fifty ways of managing them are. A unified data layer, Microsoft Fabric among others, manages and gates them in one way, on premises included.
  • Draw the line at your digital estate. Every AI solution either sits inside your tenancy, within your trust and knowledge boundary, or outside it, where someone else has access to your knowledge. Keep operational and institutional knowledge inside, and never expose it to the models.
“Knowledge protection is the key in all of this. Not the models any longer. The models will come and go.”Christoph Berlin
Pegah Zarei running a live demonstration of IntuigenceAI on Microsoft Fabric
Pegah Zarei runs the live demonstration.
04

Product talk and live demo

Ask the AI engineer for finished work, not an answer.

Pegah Zarei, VP of Product, IntuigenceAI

  • The drawing became a graph while the room watched. Added from the data lake to IntuigenceAI’s workload in Microsoft Fabric, a raw P&ID was parsed into symbols, lines and tags with every connection kept. The same drawing in a general-purpose model scores under 10 percent once 20 P&IDs are in context.
  • A team of AI engineers traced a cold column to its likely cause. A plant engineer agent routed the question to the chemical, mechanical, E&I, APC and maintenance specialists. They found the instruments and alarm limits, the two steam reboilers, and on a phone call named the flooded condensate drum starving one of them.
  • One request became a repeatable workflow. Troubleshoot the column, notify the shift engineer, raise a work order, alert below set point: planned from the graph, scheduled every six hours, shared as a template. The agents also join Microsoft Teams calls.
“The difference is accuracy, latency, cost, and finished work as opposed to “Chat AI”. You don’t just ask it questions, you ask it to perform the work.”Pegah Zarei
Hany Soliman of Microsoft presenting on industrial AI at scale
Hany Soliman on industrial AI at scale.
05

Keynote: Physical AI outcomes

The race is not who can build it. It is how to get to value first.

Hany Soliman, General Manager, Energy and Resources, Microsoft

  • Chatting with your data was phase one. Phase two is the income statement. That only happens when a workflow runs end to end across departments and the organization flattens around it. One company put one person from each system in a room and asked how a startup would serve its customers, then reorganized around the answer.
  • Pilot paralysis is fragmentation in disguise. Each pilot proves something inside one department’s data and application, so the value stops at its edge. The magic starts when the departments and their data sets work together.
  • Index the data where it sits. Do not lift and shift it. By the time a data lake was aggregated and contextualized, the data was obsolete. Pointing at, mirroring and shortcutting data in place starts a use case in hours. One operator cut 500,000 tons of emissions by putting 250 billion signals to work on its engines.
“Don’t go back and let your teams play with it. Run a hackathon, ask them to reimagine a broken process, and have the CEO give the awards.”Hany Soliman
Adam Bogobowicz of Microsoft presenting a 21 times reduction in tokens
Adam Bogobowicz on the economics of AI at scale.
06

Keynote: Make your tokens do more for less in industrial AI

AI cost multiplies. Plan it, control it, optimize it.

Adam Bogobowicz, Senior Director, Microsoft

  • Four levers multiply, and each can move tenfold. Interactions, agent steps, cost per step and token price. Cloud FinOps fights for 10 to 20 percent; with AI the swing is orders of magnitude, so a plan built from a pilot can be off by ten times.
  • A monthly budget is not a control. On a quiet Sunday an agent retrying a failed historian call every ten seconds burned 1.3 billion tokens before anyone read the alert. A circuit breaker that trips within 30 minutes beats a monthly budget by about 1,500 times.
  • Spend where a miss is expensive. Route everything else. Grounding a question in a knowledge graph cut it from 150,000 tokens to 7,000. Add model routing and the gap reaches 42 times. Pay for the best model on safety and resiliency calls, and send routine questions to a good-enough one.
“If you do the planning wrong you can be off by ten times. If you do control wrong, one Sunday afternoon can ruin your budget. If you don’t optimize, your budget can be fifty times higher than it needed to be.”Adam Bogobowicz
Cover of The Definitive Guide to Industrial Agentic AI and Microsoft Fabric/Copilot Integration Reference Architecture

The handbook, coming next week

The Definitive Guide to Industrial Agentic AI

Eleven chapters on putting industrial AI to work across your plants, with the Microsoft Fabric and Copilot integration reference architecture. The digital edition goes out next week. Email us and you will be among the first to receive it.

  1. 01Industrial AI reference architecture for Microsoft Fabric and Copilot
  2. 02Sovereign industrial AI for plant data, models and know-how
  3. 03Onboarding your plant’s data in 48 hours
  4. 04Optimizing compute and storage for industrial AI
  5. 05Industrial knowledge graphs as infrastructure for AI
  6. 06Your next engineer isn’t human
  7. 07Capturing your retiring experts’ judgment
  8. 08Capturing tacit knowledge in industrial operations
  9. 09Conversational industrial AI agents for field workers
  10. 10Data security in industrial AI
  11. 11Industrial data readiness for reliable AI deployment
Send me the guide
Guests at The Industrial AI Summit in Houston during an afternoon keynote

What comes next

The Industrial AI Summit returns to Houston, and to cities around the world.

The summit has gathered process-industry leaders in Tokyo, Seoul and now Houston. IntuigenceAI will be at the next edition as a partner. Register, and come and find us there.

Register for the next summit

Start on your plant

Figure from The Definitive Guide to Industrial Agentic AI01

Executive Briefing

A 60-minute private session on how plant-knowledge digitization fits your digital roadmap.

Figure from The Definitive Guide to Industrial Agentic AI02

Plant Knowledge Mapping Workshop

A half day reviewing your documents, systems and workflows, returning a digitization plan.

Figure from The Definitive Guide to Industrial Agentic AI03

Pilot Design Sprint

A two-week scoped engagement to define the first proof of concept on your plant.

Figure from The Definitive Guide to Industrial Agentic AI04

Sandbox Access

A 60-day trial environment with sample datasets and a named forward-deployed engineer.

Book an Executive Briefing Book a demo