Construction technology vendor Autodesk promised more investment in artificial intelligence workflows and data sharing at its annual Autodesk University user conference in San Diego. The company also detailed a future where granular data will flow from Autodesk authoring applications such as Revit, AutoCAD and Civil3D onto cloud platforms such as Autodesk Docs and Forma and also from partner companies such as geospatial information systems provider Esri.

Autodesk announced October 15 that Forma is now connected to Revit using Autodesk Docs via a translation connector. Amy Bunszel, Autodesk’s executive vice president of AEC, said that Docs is being set up as the company's AECO data repository where all project data can be connected, organized, and secured in a proprietary cloud-hosted database that can be accessed via Autodesk APIs and partner applications. She also said the company is now running a private beta for an API that gives access to Revit geometry, the actual parametric data of a design, free of files. Bunszel said project team members can access Revit model data via API, not requiring a file exchange such as in an RVT or IFC file, in the beta.

Forma Takes Shape

CEO Andrew Anagnost said that Autodesk' conceptual design tool Forma, which began its development life as Spacemaker.ai, is "a different paradigm"  for design and that the general advance of large language models has made users more comfortable with the natural language prompts and chat that software like Forma will use in the future.

He added that Forma is where Autodesk sees design starting in the future, answering big questions up front at the conceptual stage. While Anagnost said that Forma already has sustainability analysis at that early level, Autodesk plans to take it more into design and digital-twin analysis roles as the platform matures.

"You're going to see it merging with geospatial data, that's just a natural thing for getting a whole conceptual suite," said Anagnost. "It's going to start moving downstream  into things that Revit classically does as well. It doesn't mean it has to swallow all of Revit, and you know that would take a long time, but it can certainly do things that Revit does today."

Anagnost reiterated that Revit is not moving entirely to the cloud nor will its design functions go away as a standalone authoring tool. Rather, he said, its granular data, such as geometric data, is gaining the ability to flow to other platforms such as Forma and Docs. This has been a longtime request from users, that they gain access to granular data, so they can break down geometry by selection and only share what's needed rather than full 3D model files.This followed the Revit data model API released in June. 

"You're just seeing the direction that we're taking Forma, just maturing it, and it is a new paradigm, right? We're not trying to do Revit in the cloud, which is not a good way to take in the world, especially in the world of AI," Anagnost says. "You actually want to work in a way where AI, and Forma has been AI for a long time and it natively started that way, would want to move forward with AI in a way that actually enhances the productivity of the people that use these tools."

A Civil Exchange

For civil engineers, the company announced a granular data exchange via an expansion of its existing partnership with GIS data provider Esri. The integration introduces Esri’s geospatial reference data from its flagship ArcGIS mapping technology into Forma for early design and planning stages. Micah Callough, Esri's technical director for AEC and a former Arcadis vice president, said that the integration is already being tested by users and will be available in early 2025. The integration, in theory, would give users the ability to bring Esri map layers and attributes into Forma for those early design decisions and, depending on the ease of data transfer, could eliminate rework and data restructuring that has fallen on the shoulders of engineers for years. 

Earlier this year, Esri’s ArcGIS basemaps were integrated with Civil3D and AutoCAD providing detailed geospatial data and mapping capabilities. The integration of ArcGIS data into Autodesk Forma works toward the two companies' strategic alliance to unify GIS and BIM. The first product to come out of the alliance was ArcGIS GeoBIM, a translation tool that allowed users to visualize and move data between the two systems. Both companies later added connectors for InfraWorks. 

Translating data from GIS to BIM and back has been a longtime pain point for civil engineers. Autodesk and Esri's press release promised it would allow users to access GIS data quickly in BIM and Esri’s ArcGIS basemaps and select ArcGIS Living Atlas of the World data layers would all be included in the integration.

Openness vs. Intellectual Property

Last week, Bentley Systems continued to tout its commitment to open project data exchange through applications such as recent visualization acquisition Cesium. Autodesk has previously used similar language but took a different tone at this conference. While continuing to say that proprietary file formats are a thing of the past and data needs to flow freely among AEC project stakeholders, Anagnost and other Autodesk executives laid out workflows wherein data could flow through Autodesk's APIs, but also articulated how it would flow best through partner companies such as Esri and Microsoft and into clouds that had integrations developed via strategic alliances. Cesium itself announced a Revit add-in October 15. 

On the AI front, Autodesk's investment in Project Bernini, an experiment in training an AI to create 3D content that can be used by professional engineers, is a long-term play to better understand how to train AIs to create 3D, geospatial design imagery, both Anagnost and Bunszel said.

"The data used to train [Project Bernini] is just as important as the method you use to train it," Anagnost says. "You have to teach the computer to speak a certain language. Why large language models fail on so many things is because it's actually coded word by word, and it doesn't actually necessarily understand relationships between words. People run into these things all the time with large language models, we're actually creating training methods that understand 3D geometry in a deep way. They don't create pictures. They actually create real geoemetry and those training methods are data independent."

Anagnost said that to create better AIs that can be used for construction and manufacturing, you need the methods, the data and the computing power, a tall order for any company looking to create better large language models. 

"It's a lot of things to be able to have at once, but we do like to support the research community, because we do want to make some of these problems, these hard problems, easier for everyone," he says. Anagnost added that research is something Autodesk is committed to for the greater design community but also for the innovations that could come from efforts like Project Bernini. "That's something we'd like to participate in. And, yes, we also like to keep our secrets, we're a commercial company. Who wouldn't? Why would we produce something innovative and smart and not protect it?"