On September 11, 2026, FANUC announced the development of an artificial intelligence agent designed for welding robots. It can read product drawings and understand the shape and materials of a workpiece, then automatically generate welding conditions and the robot’s movement path. The company aims to reduce the manual work required to convert an engineering drawing into an executable program for an industrial robot.
The system relies on Google Cloud technologies. According to the announced presentation, an operator can give instructions to the artificial intelligence agent so that it analyzes the drawing and suggests welding process settings and robot movements before executing them on the system. FANUC plans to showcase the technology at the 2026 International Welding Show, scheduled to take place from September 16 to 19, 2026, and to begin rolling it out in December of the same year.
How does the system work?
The system reads drawings of parts used in manufacturing and identifies the information needed for the process, such as the shape of the workpiece and the materials used to manufacture it. It then creates data related to the welding process, including operating conditions and robot movements. This approach allows the operator to move from the drawing to robot setup using natural-language instructions instead of building the entire program manually.
The drawing-reading process uses a camera integrated into the teach pendant or operating panel of the FANUC CRX robot. Another camera can also be added when needed, allowing the system to accommodate differences in the usage environment or the method used to input drawings.
What changes in practice?
The main practical value of the announcement is not the addition of a new welding function to the robot, but the transfer of part of the programming and engineering-understanding work to an artificial intelligence agent. This could help operators set up welding tasks more quickly, especially when drawings change or tasks are repeated with different workpieces. However, the source provides no figures on setup time or the success rate of the programs generated by the system, so the extent of the improvement compared with conventional programming cannot yet be measured.
FANUC says that users do not need to create a separate Google Cloud account to use the system, as access to the agent is expected to be provided through a subscription mechanism linked to FANUC. The company also indicates that the system does not rely on an external power source for the connected FANUC robot, according to the information included in the announcement.
Open data and limitations
The system includes the use of Google Cloud’s Gemini Enterprise for aspects related to artificial intelligence. FANUC explains that the product drawings read by the system are not used to train the artificial intelligence for the benefit of other users. However, the announcement does not provide sufficient details about the mechanisms used to verify generated programs, how incomplete drawings are handled, or the extent of operator intervention required before the robot is operated.
From certi.news’s perspective, the announcement represents an important trend at the intersection of artificial intelligence and industrial robots because it directly connects artificial intelligence models to the manufacturing-process setup stage. However, the transition from a demonstration to broad production use will depend on the accuracy of drawing interpretation, the safety of generated commands, and operators’ ability to review the results before operating the robot.