"For physical AI to work in industrial settings, it has to support not only recognition of the physical environment but also causes and effects, reasoning and industrial context."

Pascal Daloz (파스칼 달로즈), CEO of Dassault Systemes, said physical AI, an AI paradigm drawing attention after large language models, is bound to face limits without reasoning capabilities that understand context in the real world. He presented an Industry World Model and virtual companions based on it as the solution.

After a keynote speech at the 3DEXPERIENCE World 2026 conference held from Feb. 1 to 4, Daloz said at a press briefing that the Industry World Model blends perception and reasoning about reality. He said it is not a simple chatbot but will enable companion-grade virtual companions that connect engineers, designers and business personnel.

At 3DEXPERIENCE World 2026, Dassault Systemes holds its annual conference for users and partners of its SolidWorks 3D solution product line. With 6,000 people attending this year, Daloz focused on how the Industry World Model and virtual companions are changing work in the industry and what changes they will bring going forward.

He said virtual companions are on a different level from simple AI chatbots. They are AI that can literally serve as companions. When design and a virtual companion are combined, users can design models in natural language, and they can also cover error analysis and simulation with AI while following governance. Dassault Systemes demonstrated multiple times at the event 3D modelling conducted through a natural-language interface. It is also expanding the product line that supports it.

It unveiled a beta version of the virtual companion Aura at last year’s 3DEXPERIENCE World. At this year’s event, it also released preview versions of Leo and Marie, with an official launch planned in the first half of the year.

The company said that unlike Aura, which supports basic work, Leo and Marie support high-end engineering as well as 3D modelling. It said they play a role similar to an engineer handling physical-based analysis across structure, mechanical, simulation and manufacturing. Marie handles more scientific aspects such as materials in the design process.

Daloz said Aura is a specialised adviser for strategy, business and compliance, while Leo can be seen as a hands-on engineer focused on manufacturability and structural validity. He said Marie plays the role of a researcher supporting theoretical verification during design based on scientific knowledge. He said the virtual companion will not be one, but three AIs each taking on a role to deliver true knowledge-based collaboration.

He said the output produced by a virtual companion is not at the level of simply carrying out commands. If a user enters, "Design a steel frame to support a water tank," the AI generates a full model within minutes, including structural analysis. Users can build advanced models through simple conversation without repetitive work or using complex functions.

It can generate 3D models directly from PDFs or images, or automatically convert existing supplier CAD files into a parametric model, and it can be modified if needed. If an error occurs, the AI analyses the cause and informs the user. An 'Assembly Doctor' function that even presents the causes and solutions when assembly performance deteriorates is also close to being deployed in real operations.

Daloz said a differentiator in trust is that users can control the AI when needed. He said drawings can be generated automatically, but standards for drawings and view configurations can also be customised by the user. He said enterprise functions such as project management and change approval have reached a level where natural-language instructions such as "Ask my boss for approval" are possible. He said Leo and Marie will be launched to the market in the middle of this year and that this has enabled Dassault Systemes to provide an integrated solution of a new dimension.

The message AI is sending to industrial design is clear. Daloz said AI does not replace physical constraints but works by accepting them and solving knowledge and know-how faster, better and differently. He said that is the essence of a virtual twin.

Daloz said he sees the convergence of AI in industry as a turning point toward a "generative economy". He said that until 30 years ago, industry focused on making things, but it has now shifted to a structure that creates "knowledge and know-how". He said AI is a tool that amplifies this, and in the process products increasingly become software-defined. He said the structure is changing so that cars, molecules, systems and even business models are defined by software.