Key Takeaways
- If your machining knowledge lives in a retiring machinist's head, that is a data problem before it is an AI problem. Start capturing.
- Ask every industrial AI vendor where the human sign-off sits. If the answer is vague, the demo is the product.
- The BOM-sync and trustworthy-data interviews are the least glamorous and the most important. AI on bad product data is fast bad product data.
Short Answer
At IMTS 2026 in Chicago (September 14 to 19, 2026) Michael Finocchiaro recorded 23 interviews on industrial AI, sponsored by Aras. They cluster into five themes: capturing machinist knowledge and automating CNC programming (Neuramill, Limitless Labs, Eureka, Manukai, Hexagon), physical AI and robotics (Teradyne Robotics, Flexxbotics, Bardin AI, Thrixel), engineering AI and CAD (InfinitForm, PTC Onshape, C-Infinity, CoLab, Cosmon, Makistry, Dirac), data and the digital thread (Capvidia, IQBrain, MachineMetrics, Lambda Function), and the platform view from Siemens DISW and Tulip. Every interview is on the IMTS 2026 Full Interviews playlist on the DemystifyingPLM YouTube channel.
- The most repeated theme was not AI capability but knowledge capture. Five interviews were about what happens to shop-floor expertise when the people who hold it retire.
- Every CNC-adjacent conversation landed on the same guardrail. AI can propose the toolpath; simulation and a human sign-off decide whether it runs.
- Physical AI was on the floor, not on the roadmap. Teradyne, Flexxbotics and Bardin were talking about deployments and the engineering bottleneck in front of them, not demos.
- The CAD conversations split cleanly. Text-to-CAD has a geometry problem (Onshape), generative design needs parametric output (InfinitForm), and AI-generated geometry needs a validator (C-Infinity).
- Siemens and Tulip both framed agents the same way. Agents that act, with an engineer making the call, and the skills gap as the reason the timing is now.
Four days at McCormick Place, 23 conversations, and one question that came up in nearly all of them, asked twenty different ways: when the machinists retire and the AI shows up, who is actually in control?
This is the index of the full IMTS 2026 interview series, recorded September 14 to 19, 2026 in Chicago and sponsored by Aras. Short highlights from each conversation are on the Best Moments playlist. Below, the interviews grouped by theme, with the guest, the one question the conversation turns on, and the buyer's guide it feeds.
Machinist knowledge and CNC programming
The show floor is machine tools, so it is no surprise that the deepest cluster was about the people who program them and what happens when they leave.
Airplane Orders Pile Up. Expert Machinists Retire.Nick KhormaeiNeuramillAircraft demand is rising while experienced machinists retire. High-precision manufacturing with people kept in control.Watch on YouTube (opens in a new tab)
CNC Programming Is Ready for an AI UpgradeDavid PrievLimitless LabsProgramming as the bottleneck, and what an AI copilot for CNC actually captures.Watch on YouTube (opens in a new tab)
AI Can Crash Your CNC Machine. Simulation Comes First.Shaun MymudesEurekaThe guardrail every CNC-adjacent conversation landed on: nothing AI-generated runs without simulation.Watch on YouTube (opens in a new tab)
When Machinists Retire, Where Does Their Knowledge Go?Daniel WälchliManukaiCapturing process knowledge before it walks out of the door.Watch on YouTube (opens in a new tab)
Your Shop-Floor Know-How Is Your Competitive AdvantageHiren KumbhojkarHexagonManufacturing know-how and the data that has to sit underneath industrial AI.Watch on YouTube (opens in a new tab)
These five map directly onto the AI machining layer in the Best CAM Software 2026 guide, where the SWARF framework treats postprocessor and simulation reality as the dimension buyers underweight most.
Physical AI and robotics
Physical AI Is Already HereJean-Pierre HathoutTeradyne RoboticsDeployments, not demos.Watch on YouTube (opens in a new tab)
The Future Is Not a Lights-Out FactoryTyler BouchardFlexxboticsRobot orchestration around people rather than instead of them.Watch on YouTube (opens in a new tab)
Automation Has an Engineering BottleneckFay GoldsteinBardin AIWhat stands between an automation demo and a working deployment.Watch on YouTube (opens in a new tab)
Rearrange Your Factory Without Moving a MachineRana HanockaThrixelSimulating layout changes before anyone touches a forklift.Watch on YouTube (opens in a new tab)
Flexxbotics sits in the automation orchestration row of the Best MES Software 2026 MINT scorecard, and the layout conversation belongs next to the digital factory strand in What is Manufacturing Process Planning?.
Engineering AI and CAD
The CAD conversations split cleanly into three positions, and they disagree with each other in useful ways.
Text-to-CAD Has a Geometry ProblemDavid Anderson and Michael LaFlechePTC OnshapeWhy prompting a B-rep kernel is harder than prompting a renderer.Watch on YouTube (opens in a new tab)
Engineers Shouldn't Just Be DraftersMichael BogolomnyInfinitFormManufacturing-aware generative design that outputs a parametric feature tree.Watch on YouTube (opens in a new tab)
AI Can Generate It. Who Will Validate It?Sai NelaturiC-InfinityThe validation gap behind every generated geometry.Watch on YouTube (opens in a new tab)
Your Design Knowledge Is Stuck in NotebooksTaylor Young and Jeremy AndrewsCoLabDesign review knowledge as an asset nobody has been storing.Watch on YouTube (opens in a new tab)
When Engineers Become Editors of AI AgentsRui AguiarCosmonWhat the engineering job looks like when the agent drafts and the engineer edits.Watch on YouTube (opens in a new tab)
AI Takes Action. Engineers Make the Call.Parth MehtaMakistryAgency with a human decision at the end of it.Watch on YouTube (opens in a new tab)
Your CAD Changed. Did Your Instructions?Ben BrakenwagenDiracWork instructions that follow the model instead of lagging it.Watch on YouTube (opens in a new tab)
The geometry argument is the one this site has spent the most time on. Start with What is a Geometric Kernel? for why text-to-CAD is hard, and the Best CAD Software 2026 guide for where InfinitForm and Onshape sit in the landscape.
Data, BOMs and the digital thread
The least glamorous interviews and, if you run a plant, the most important.
Your Digital Thread Starts With Trustworthy DataFrank PopielasCapvidiaModel-based definition and data you can actually trust downstream.Watch on YouTube (opens in a new tab)
When Your BOMs Fall Out of SyncSoufiane ElaamiliIQBrainThe EBOM-to-MBOM gap, seen from the AI side.Watch on YouTube (opens in a new tab)
Your Factory Has Data. Why Are Margins Still Disappearing?Bill BitherMachineMetricsDashboards show what happened. What helps a manufacturer decide what to do next?Watch on YouTube (opens in a new tab)
AI Agents Shouldn't Mean Rebuilding Your Factory SoftwareTanmay AggarwalLambda FunctionAgents on top of the stack you already have.Watch on YouTube (opens in a new tab)
For the BOM conversation, EBOM vs MBOM is the background reading. For the trustworthy-data one, What is Supply Chain Traceability? and What is CAD Interoperability?.
The platform view
AI Agents Don't Just Look. They Take Action.Rahul GargSiemens DISWWatch on YouTube (opens in a new tab)
The Skills Gap Is Already Here. What Can Industrial AI Do?Chris PollackSiemens DISWWatch on YouTube (opens in a new tab)
What AI Changes on the Factory FloorNatan LinderTulip InterfacesWatch on YouTube (opens in a new tab)
Siemens and Tulip framed the moment the same way from opposite ends of the market: agents that act, with an engineer making the call, and the skills gap as the reason the timing is now rather than in five years. Tulip is scored as a composable platform in the MES guide; the Siemens portfolio runs through the Siemens Spotlight.
What I took home
Three things. The knowledge-capture theme was louder than the AI theme; five founders built companies around a retirement wave, not a model. The guardrail question, where the human sign-off sits, separated the serious vendors from the demos within about ninety seconds of each conversation. And the companies solving the boring problems, BOM sync and trustworthy data, are the ones the others quietly depend on.
Every company in this series is profiled on ThreadMoat, with funding, positioning and the ThreadMoat Strategic Disruption Potential score where one has been assigned. The ICC 2026 interviews, recorded the following week in Sacramento, pick up the same questions from the automation side.
Want to listen instead of read? 56 DemystifyingPLM articles are available as audio.
Browse audio →Looking up PLM terminology? Browse the canonical reference.
PLM Glossary →Cite this article
Finocchiaro, Michael. “IMTS 2026: 23 Interviews on Industrial AI from the Show Floor.” DemystifyingPLM, September 26, 2026, https://www.demystifyingplm.com/imts-2026-interviews
Michael Finocchiaro (opens in a new tab)
PLM industry analyst · 35+ years at IBM, HP, PTC, Dassault Systèmes
Firsthand knowledge of the evolution from early 3D modeling kernels to today's cloud-native platforms and agentic AI — the history, strategy, and future of PLM.
