By Randall Scott Newton, Managing Editor
and Vektor, AI Agent Reporter
If you feel like your engineering software has been shouting “AI” at you for years without much changing, you aren’t alone. We have survived the rise of copilots, those helpful but limited chatbots that can summarize a meeting or write a simple script if you ask nicely.
In 2026, the industry is moving toward something fundamentally different: AI agents. To understand the shift, think of each as a GPS. The copilot provides a route (“optimize the bracket”) and warns about traffic (“check the weight”), but the driver (engineer) is still manually executing every turn (running simulations, changing the model). The agent, by contrast, can be told the destination (“optimize this bracket for weight and 3D printing”). It researches the material, runs the simulation, fixes the geometry, and presents you with three verified options. We aren’t there yet, but the shift is happening rapidly.
The leading CAD/CAE/PLM vendors have moved from talking about roadmaps to shipping actual agentic features this year. Here are snapshots of what’s changing.
Autodesk the connected assistant
Autodesk is focusing on embedded AI agents living inside Fusion and Revit. Instead of a separate window, these agents will become a practical layer that handles the busywork.
Their agents will read drawing sets and automatically generate a Bill of Materials or flag schedule risks in a construction project before they happen. The goal is to reduce the click-heavy administrative parts of design.
PTC the digital thread specialist
PTC is taking a very disciplined approach with its Windchill AI for PLM. They aren’t just making a chatbot; they are building agents that apply intelligence across the entire product history.
The Parts Rationalization agent will scan the entire enterprise and say, “Hey, you’re about to design a new bolt, but we already have a nearly identical one in the warehouse. Use that instead.” This one step alone could stop “duplicate part” bloat and save millions in inventory costs.
Siemens the industrial teammate
Siemens is bringing AI onto the shop floor with new Collaborative Intelligence features; the AI acts as an engineer’s digital teammate. These new agents won’t just write code, they will also troubleshoot machines in real time by reading error codes and comparing them to 30 years of maintenance manuals. The goal is to solve the skilled labor gap by making sure every junior operator has the knowledge of a 30-year veteran.
Synopsys/Ansys from silicon to systems
Following the 2025 acquisition of Ansys, the combination quickly moved to make sure their expanded product line works from nanometers to meters. New AI agents will simulate electronics and physics. This means an agent can optimize a circuit board while also checking how the heat from those chips will warp the plastic casing. The goal is faster, cheaper prototyping for complex software-defined products like electric vehicles.
Dassault Systèmes the IP guardian
Dassault Systemès is extending its 3DEXPERIENCE platform to embrace AI agents. AURA is an AI companion accessible in 3DSwym and from within SOLIDWORKS (MySession). It can answer questions using information stored on the 3DEXPERIENCE platform while emphasizing IP protection. In SOLIDWORKS itself, near-term “assistive AI” includes workflow prediction (Command Predictor) and AI-assisted drawing generation, aimed at speeding up detailing and documentation. At the recent 3DEXPERIENCE World 2026 the company described an expanded lineup of SOLIDWORKS “virtual companion” AI agents.
The end user impact
Agents don’t replace engineering judgment; they act as high-speed “research and execution” assistants. The human engineer must still sign off on the final design.
Engineering will shift from doing the calculation to verifying AI agent work is correct.
AI agents are only as good as the company’s data. If PLM or CAD files are a mess, the agent will only make mistakes happen faster.
[AI/Human contribution: This article was written by a human; an AI large language model with specific agency did the initial research. Sometimes the text presented was good enough for direct use in the article. There were several rounds of guidance prompting, and a final round of human editing to finish the article. As a result, the exact word-for-word “who or what wrote this” is a mishmash.]
Your comments are welcome