en.Wedoany.com Reported - On June 9, Seth DeLand, Product Marketing Manager at U.S. engineering software company MathWorks, stated that agentic AI is opening up larger problem spaces for engineers. His insights, focused on MATLAB, Simulink, model-based design, simulation verification, and automated task execution, reflect how engineering software companies are advancing generative AI capabilities from code assistance to executable and verifiable engineering processes.
The key change with agentic AI lies in "iterative execution." After engineers set task goals and success criteria, the AI agent can continuously iterate on code writing, model invocation, simulation runs, error correction, and result checking.
In traditional generative AI-assisted engineering scenarios, engineers typically input problems into a chat interface, copy the generated code into MATLAB or another development environment to run it, and upon encountering errors, return to the chat interface to describe the problem, receive modification suggestions, and continue testing. Recent materials released by MathWorks on agentic AI for MATLAB and Simulink show that agentic AI connects large language models with engineering computation, simulation environments, and local data through the Model Context Protocol and toolkits. This enables AI agents to call MATLAB functions, run Simulink models, read workspace data, execute code, analyze errors, and continue refining results. This model pushes AI from "giving suggestions" to "executing tasks," shifting engineers' focus toward goal setting, constraint definition, model validation, and result review. For scenarios such as control systems, embedded software, signal processing, mechanical design, autonomous driving, robotics, and industrial equipment R&D, if AI agents can operate within existing engineering toolchains, they can take on more repetitive experiments, parameter adjustments, model connections, and report generation, allowing engineering teams to concentrate on system architecture, boundary conditions, safety verification, and trade-off decisions.
MathWorks' related technical materials also emphasize that reliability and traceability remain core issues as agentic AI enters engineering workflows. Engineering development cannot rely on "intuitive coding"; embedded systems, controllers, and industrial software all require clear requirements, reuse of validated models and toolboxes, and confirmation of results through simulation and phased testing.
This assessment has a direct impact on the engineering software market. MATLAB and Simulink have long served the automotive, aerospace, communications, energy, industrial automation, and scientific research sectors, with many customers' R&D processes relying on model-based design, simulation, code generation, and test verification. If agentic AI remains confined to general-purpose chat interfaces, it will struggle to enter highly constrained engineering environments. However, when AI agents can access professional tools, understand model structures, invoke simulation environments, and output verifiable results, the value of engineering software platforms will be further amplified. Software vendors need to provide interfaces, domain expertise, toolchain connections, and permission controls that AI agents can call upon, while corporate R&D teams must redesign human-machine collaboration workflows, combining AI generation, automated execution, engineer review, and quality closure.
The implementation of agentic AI in engineering will also reshape the skill structure of engineers. Engineers will still be responsible for system definition, physical mechanism judgment, test plan design, compliance boundary confirmation, and ultimate accountability, but some basic coding, batch experiments, data organization, and iterative verification can be assisted by AI agents. For complex product R&D, this means teams can explore more design options within the same timeframe, shorten the path from conceptual modeling to simulation verification, and identify issues in parameters, models, and control strategies early on. Subsequent variables focus on enterprise data security, model permission management, multi-toolchain adaptation, generated result verification, and engineering responsibility boundaries. Only when clear rules are established in these areas can agentic AI transition from a demonstration tool to a scaled engineering production process.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
