We’re on the threshold of a fundamental shift in how humans interact with software: from operating applications to co-creating with agents.
Developing a product, resolving an engineering challenge or investigating a scientific hypothesis requires a sequence of decisions and actions. Requirements evolve. Simulations reveal new constraints, and each result drives what happens next.
AURA, LEO and MARIE, Dassault Systèmes’ autonomous, long-running and self-evolving Virtual Companions, are designed for this way of working. Operating on the 3DEXPERIENCE agentic platform, they reason, plan and act across evolving missions, co-creating and co-operating with users to address complex industrial challenges.
Virtual Companions transform the intelligence of Industry World Models into actions. Each brings complementary expertise:
- AURA, the business expert, co-creates with business leaders and project managers to turn business objectives into executable plans and measurable outcomes.
- LEO, the engineer, co-engineers with designers and simulation experts to turn ideas into manufacturable, high-performance products.
- MARIE, the scientist, co-creates with scientists and researchers to turn scientific hypotheses into discoveries and innovations
From Answering Questions to Carrying Out Missions
The difference between a conversational assistant and a Virtual Companion is not simply how well it answers a question. It is how it turns an objective into an outcome.
Each Virtual Companion brings competencies—a structured set of skills that guide users, reveal insights and execute relevant work across business, product design, engineering, simulation, manufacturing and operations.
By co-creating with a Virtual Companion, users extend their capabilities, explore more alternatives in less time and turn ideas into actionable outcomes.
What Makes a Virtual Companion a Long-Running Agent?
“Long-running” describes more than the duration of an agent’s execution. It describes its ability to maintain continuity across a mission in a changing industrial environment —reconciling the objective, the plan and the results as the work unfolds.
To carry that mission forward, autonomous Virtual Companions rely on five defining characteristics:
- Mission-driven execution. Virtual Companions work toward an objective, rather than waiting for a separate instruction for every task.
- Planning and orchestration. They translate high-level objectives into an executable plan and coordinate the relevant skills, agents and human experts.
- Adaptation. They update the plan in response to execution results or changes in the industrial context.
- Human oversight. They involve users at meaningful decision points to review alternatives, validate outcomes and redirect the work, preserving human accountability.
- Knowledge capitalization. They capture interactions, corrections and completed work, then leverage that experience in future missions to continuously refine how they work —making them self-evolving.
Industrial Intelligence Grounded in Science
To act effectively over time, an agent must understand its industrial context.
Dassault Systèmes’ Industry World Models provide that industrial intelligence. They combine:
- Science: Physics, the laws of physics and their application across disciplines, such as thermodynamics, fluid mechanics, mechanics, electromagnetism and aerodynamics … but also chemistry, materials science and biology.
- More than 40 years of industrial knowledge and know-how developed with our customers.
- Reasoning: This ability to understand objectives, activate relevant knowledge, orchestrate actions and generate solutions.
The 3DEXPERIENCE Agentic Platform: Orchestrating Intelligence and Execution
The 3DEXPERIENCE platform is becoming an Agent-as-a-Service platform with the 3DEXPERIENCE agentic platform. It provides a secure environment for orchestrating our Virtual Companions, leveraging NVIDIA technologies.
The 3DEXPERIENCE agentic platform, relying on the sovereign OUTSCALE AI Factories is the place where Industry World Models are operated and Virtual Companion competencies are orchestrated.
The 3DEXPERIENCE agentic platform enables horizontal openness by design. Through open agentic standards (A2A), it supports collaboration with third-party agents across enterprise systems, including ERP, CRM and other PLM platforms. Industrial missions can therefore be coordinated across a broader enterprise environment, rather than remaining confined to a single application.
Accelerating Supporting Long-Running Industrial Agents with NVIDIA Technologies
Dassault Systèmes’ collaboration with NVIDIA brings complementary technologies to this architecture: reasoning models, agent optimization and controlled execution.
Dassault Systèmes combines NVIDIA Nemotron open models with its Industry World Models to connect AI reasoning with deep industrial context.
NVIDIA NeMo enables continuous agent specialization, optimization and evaluation—values that matter for repeatable industrial execution.
Dassault Systèmes is also using NVIDIA OpenShell, a secure runtime which governs long-running autonomous agents. This control helps support sustained execution within defined boundaries and is part of the new NVIDIA OpenShell Reference Architecture.
These technologies play distinct roles. NVIDIA contributes models, optimization capabilities, execution controls and accelerated tools. Dassault Systèmes contributes the science-grounded representations, industry driven knowledge, customer know-how and agentic orchestration that make them relevant to industrial missions.
Related Resources on Industrial AI
Nicolas CerisierNicolas Cerisier is vice president, 3DEXPERIENCE Platform R&D at Dassault Systèmes. In this role, he is leading transformation of the 3DEXPERIENCE platform to an Agent-as-a-Service platform.
Hubert MassonHubert Masson is the 3DEXPERIENCE Platform 3DSwym Senior Director at Dassault Systèmes.