I attended AU 2026 a few weeks ago. This is an article in my series of posts about what I learned there. If you missed my previous articles about Autodesk University, please check them here –
AU 2026: MCP Connects the Tools but Who Connects the Product Context
Invisible PLM and the Engineering Controls That Still Matter
AU 2026: The Platform Question Is No Longer APIs. It Is Context.
AU 2026 and the Autodesk Bet on Context for PLM
The lawnmower factory and airport demonstrations at AU 2026 extend the digital thread into operations. The next test is whether operating experience changes engineering decisions and preserves the evidence behind what the organization learns.
Autodesk made the biggest acquisition ever buying MantainX for $3.6B in cash and advancing unified platform in operations.
Autodesk’s strategy is to converge design, make, and operate workflows, ensuring data and insights flow seamlessly in a continuous lifecycle. With the creation of Autodesk Operations Solutions (AOS), the company is bringing together its operations capabilities under a unified platform.
“Operations is where organizations manage the systems, assets, facilities, and workflows that keep their businesses running every day,” said Stephen Hooper, SVP of Autodesk Operations Solutions. “Autodesk is enabling customers to move from managing operations to continuously improving them, deriving more value from their data, and positioning them for the coming wave of AI-driven workflows. MaintainX brings deep expertise in maintenance and frontline operational workflows that complements this broader strategy.”
The operations segments at AU 2026 brought the lifecycle discussion into a place where engineering decisions face a longer test. A factory has to produce at the required rate. An airport has to keep people moving. Equipment has to be maintained by teams who may never have met its designers.
That is why I paid attention to the connections between Tandem, Forma, FlexSim, and MaintainX. Autodesk is extending its design and make story into operations. For the PLM industry, the interesting question is how operating experience can become useful knowledge for the next engineering decision.
We have spoken about the digital thread for many years. The AU demonstrations provided an opportunity to examine what happens when the thread reaches a working facility, and what else is needed for the organization to learn from it.
An engineering change becomes a capacity question
Day 1 returned to the lawnmower scenario after the supply chain and engineering discussion. The production team needed to evaluate an additional machining step while increasing output from 50 to 60 lawnmowers per hour.
The demonstration started with an imperfect information environment. Part of the campus was older, with floor plans and spreadsheet data rather than uniformly clean models. The team used that information to build an operational view incrementally. Simulation then explored whether the production line could meet the target.
I liked this starting point. Many discussions of digital twins begin with an ideal model. Real organizations often need to make a decision using the information they already have. Connecting that information can be useful before the representation becomes complete.
The result shown on stage was a simulation outcome. It was not evidence that a physical production line had sustained the target. That distinction matters because the assumptions behind a simulation are part of the decision people need to evaluate.
For PLM, the broader lesson is that the consequence of a component change can reach into process capacity. The BOM and engineering model describe part of the situation. Manufacturing knowledge determines what the change means for delivery.
An airport connects built and manufactured systems
Day 2 used Hartsfield-Jackson Atlanta International Airport to show the convergence of construction and manufacturing information in operations. Buildings contain equipment and systems supplied by manufacturers. Their operating behavior depends on how those assets work together in a particular environment.
The security-checkpoint simulation made that relationship understandable. Changing screening-lane configurations could affect waiting times and equipment utilization. The choice also involved investment, so a technically attractive configuration needed to be evaluated as an operational and economic decision.
This is a useful extension of the lifecycle discussion. A manufacturer may know the design of a machine. A facility operator knows where it is installed, the demand placed on it, and the maintenance experience around it. Combining their information can reveal consequences that neither sees alone.
Autodesk’s operations announcement describes Tandem bringing operational information together with design data, process simulation through FlexSim, and maintenance workflows through MaintainX. I see these as components of a feedback process whose value depends on how well the learning returns to engineering.
Connected records need an explanation of the experience
A digital thread can connect a design revision to a built asset and its service history. Product memory adds the circumstances and reasoning needed to interpret that history.
Consider an asset that needed repeated maintenance. The event history may tell us what was replaced and when. The useful engineering questions go further. Was the installation different from the original assumption? Did the operator use the equipment under a higher duty cycle? Was a supplier substitution involved? Did an earlier team recognize the issue and accept it for a particular reason?
These are examples of the interpretation a team may need, not claims about the airport demonstration. They show why connecting records alone does not complete the learning process.
I have explored this distinction in my writing about digital thread and product memory. Operational data becomes more useful when the organization can connect it to the product situation and preserve the evidence behind its conclusions.
Factory learning and product learning are different
It is worth separating two feedback paths. Factory learning concerns how the organization makes a product: machining time, line balance, setup, material flow, and production quality. Product learning concerns how the delivered product performs in its operating environment.
Both can influence engineering, but they connect through different identities. A manufacturing issue may concern a process step and a particular product revision. A field issue may concern a serialized asset, its installed components, service changes, and operating conditions.
A connection between design and operations needs to preserve these distinctions. Otherwise, a team can draw a conclusion from an event that concerns a different configuration or a different environment.
The AU factory and airport examples help illustrate the breadth of the opportunity. They should not be treated as proof that every manufacturing and service feedback path is already implemented.
A feedback loop needs a decision that changes
I would test lifecycle learning by following one observation. An operating problem is detected. The team investigates it, records what the evidence supports, and decides whether engineering needs to act. A change may follow, or the team may decide to retain the design and change maintenance practice.
In either case, the reason should remain connected to the relevant asset, configuration, and records. A future team needs to know what was learned and whether that conclusion still applies.
This also requires care with causality. A failure after a substitution is evidence to investigate. It does not, by itself, establish that the substitute caused the failure. Product memory should preserve the finding as provisional until the investigation supports a stronger conclusion.
The ability to trace that reasoning matters across companies. Operators, equipment suppliers, manufacturers, and service partners each hold part of the knowledge. Useful lifecycle learning requires ways to share the relevant evidence while respecting who owns and can disclose it.
Autodesk’s operations direction gives PLM a practical question to answer. Can information from the working asset change the next requirement, design review, or production decision, with the reasoning still available afterward? A team that can demonstrate that sequence has done more than connect another data source. It has shown how the organization learns from what it builds.
Just my thoughts. YMMV.
Best, Oleg
Disclosure: I am co-founder and CEO of OpenBOM. My perspective on product data, collaboration, and AI is informed by that work.
