What I Brought Home From Autodesk University 2026

I've walked the halls of the Venetian in Las Vegas more times than I can count, mostly for cybersecurity conferences, so I arrived at AU 2026 last week expecting familiar territory. Autodesk made it feel new. The mobile app was excellent, the floor was easy to get around, and the whole event was built for the thing that actually matters at a conference: sitting down with someone and having a real conversation.

I had plenty of those. Here's what stuck with me.

Autodesk Context is a good start. Now let's go bigger.

The Day 1 keynote opened with Autodesk Context, which the company describes as a connected brain at the center of a project. It maps how information, people, decisions and activities relate, and it captures why a decision was made and what the situation looked like when someone made it. Later in the week, Mike Haley, who leads AI and research at Autodesk, summed up the strategy in one idea: AI is only as good as the context it's grounded in.

I found myself nodding through that whole keynote. It's the right message for the right audience. Construction professionals have heard plenty about chatbots, and it was refreshing to hear the biggest company in the room say the model isn't the hard part. The hard part is giving AI something true to reason over.

My only reaction was that it's still too small. Autodesk Context is built around a project, or a portfolio of projects. That makes sense for a company whose tools create the design data. But a project ends, and the business that delivered it keeps running. The asset you built operates for decades. The contractor moves its crews, equipment and suppliers to the next job and the one after that. If the context layer stops at the project boundary, it only ever sees one slice of the business at a time.

The real prize is a single view of the entire operation. So thank you, Autodesk, for putting this on the main stage. Now let’s go big.

The silos aren't in the model

It was obvious at AU how much of the real context lives outside design software. Operational systems grew up one problem at a time: scheduling here, maintenance there, then telematics, safety and fleet. Each one works on its own. Few of them connect in any meaningful way to ERP, workforce management, or each other. And the most important context of all isn't in a system. It's in people's heads.

One conversation drove that home. The CEO of a workforce planning company told the room that 40% of construction professionals plan to retire in the next five years. Think about what walks out the door with them. Decades of unwritten institutional knowledge. Supplier relationships built over years of phone calls. The intuition that tells a superintendent a job is about to go sideways before the schedule shows it. Localized safety expertise that exists nowhere in writing.

You can't replace that with a chatbot. You can start capturing it, and that's why building an ontology has become critical: a model of how your assets, people, systems and decisions relate, so knowledge has somewhere to live besides one person's memory. Oleg Shilovitsky, who writes about engineering software and was covering AU, put it well: “Every application will have an assistant. That is not the same as shared understanding.”

The encouraging part is how easy it has become to start. One of the panels made a point I agree with completely: every company needs a playground, a place to try things, fail fast and learn. The cost of an experiment has collapsed. Pick one operation and one nagging problem, connect the handful of systems and people involved, and see what you learn in 30 days. The companies waiting for certainty will get around to it after their best people have already retired.

I've watched this movie before

I spent a big part of my career in cybersecurity, and AU gave me déjà vu. For about a decade that industry sold dashboards, each with prettier charts, faster queries and more integrations. And breach after breach, the post-mortem read the same way. The alert fired. Somebody saw it. Nobody understood what it meant, because the log had no idea what that server did, who owned it, or whether that behavior was normal for that company.

The tool was never the bottleneck. The context was. Security tools only started earning their keep when we got serious about connecting and enriching the data so it reflected how the business actually worked.

Construction and heavy industry are running the same play now, with AI assistants standing in for the dashboards. I'd love to see us skip the decade of learning it the hard way.

What happens next

AU 2027 is already set for Las Vegas, September 14 to 16. My prediction is the conversation moves from designing and building toward operating, and the context layer goes from an architecture discussion to a line item in your budget.

As you continue to evaluate leading AI technology vendors, ask these four questions:

1. What does it understand that my team couldn't before? Speed is easy to demo. The real test is whether it helps your people reach a better decision than they could have reached without it.
2. Can it assess impact, or only retrieve information? Finding the right document is useful. Knowing that a late delivery affects three crews, a crane rental and a concrete pour next Thursday is what changes outcomes.
3. Can it write the decision back? If the AI recommends, a person decides, and nothing records why, the system never gets smarter. The reasoning behind decisions, including the times someone overrode the AI, is exactly the knowledge you're trying to keep.
4. Who can change what, and how do I undo it? Once AI starts acting instead of advising, permissions and reversibility matter more than the interface. You want to know exactly what an agent can touch, what it did, and how to reverse it.

This is the work we're doing at Ontollo. We build a continuously updated operational map of your assets, people, systems and decisions that AI can reason over, across the tools you already run.

I'm curious what others saw. If you were at AU, what did you walk away thinking? And if retirements and disconnected systems are keeping you up at night, drop a comment or send me a message. I'd enjoy comparing notes.