Generative AI (GenAI) is fundamentally accelerating and reshaping software development. In automotive safety-critical applications, however, faster creation can create bottlenecks at verification, validation, and traceability.
GenAI can generate code, models, tests, and documentation faster than teams can review and qualify those outputs using manual processes. Without supporting workflows, teams risk fragile architectures, weakened traceability, unpredictable behavior, and validation effort that outpaces productivity gains. For safety-critical systems, these issues directly impact certification readiness and long-term maintainability.
Automotive teams must determine how to deploy GenAI while preserving engineering authority, system understanding, and compliance.
This white paper describes how Model-Based Design provides a structured foundation for applying GenAI in automotive system and software development. By combining GenAI with Model-Based Design workflows, engineering teams can accelerate design and implementation while maintaining traceability, verification, and system-level validation across the lifecycle.
GenAI can significantly increase productivity by generating artifacts quickly, assisting with tools, and applying broad contextual knowledge to engineering tasks. However, in safety-critical automotive development, its limitations introduce risk if not managed within a structured workflow.
A shift is underway from standalone GenAI tools to agentic AI, where AI agents execute tasks based on engineer-defined goals. These agents can interact with tools such as MATLAB® and Simulink®, extending GenAI from an assistant to an active participant in development workflows.
Key enablers of agentic AI generative large language models (LLMs) are Model Context Protocol (MCP) and skill files. MCP enables LLMs to interact with engineering tools (such as MATLAB, Simulink, and Polyspace® products), turning probabilistic reasoning into grounded, deterministic execution and leveraging the understanding of physics. Skill files are textual instructions that tell LLMs how to best accomplish specific tasks and what to avoid.
The key question is not whether GenAI is useful (it clearly is) but how to apply it so that rigor, trust, and engineering accountability scale alongside speed. Without corresponding verification, validation, and traceability workflows, faster creation can shift risk into late integration, certification, and long-term maintenance.
Model-Based Design has long been a cornerstone of automotive engineering. By representing systems as executable models, it helps teams reduce complexity, automate implementation with traceability, and validate system behavior early, before issues propagate downstream.
In the age of GenAI, Model-Based Design is not only still relevant but also enhanced by GenAI. When GenAI operates inside established Model-Based Design workflows, engineers retain control of system decisions and design intent. Executable models, physics-based simulation, deterministic automatic code generation, and verification tools create a structured environment in which AI-assisted changes can be reviewed, tested, and connected to evidence.
Model-Based Design combines tools such as MATLAB and Simulink with GenAI across the development lifecycle.
In this context, visual and systematic simulation-driven verification becomes a key advantage. Executable models and simulation results allow engineers to see system behavior, understand cause and effect across domains, and supervise AI-assisted development with clarity rather than rely solely on generated code or logs.
Accelerating Design with GenAI
Design is typically the first phase where engineers experience the benefits of GenAI. In early development, much time is spent on repetitive implementation tasks rather than on higher-level system decisions. GenAI helps reduce this burden, while engineers remain accountable for outcomes. With AI agents connected to Simulink, GenAI focuses on automation and execution and engineers define goals, constraints, and trade-offs.
Engineers can use AI agents to generate architectural model drafts from textual requirements, assemble systems using vetted libraries, and apply consistent modeling patterns. In this setup, GenAI is equally valuable when working with existing designs. Large Simulink models can be refactored to improve structure and maintainability, turning what was traditionally time-consuming, repetitive manual work into a faster process.
AI automates several often-tedious design tasks, including connecting requirements to a model.
From AI‑Assisted Design to Production Software
As development moves toward production, constraints from automotive processes and standards become non-negotiable. While GenAI can assist in algorithm development, production code must be deterministic, traceable, certifiable, and real-time ready.
A practical separation emerges: GenAI supports creative acceleration; Model-Based Design enforces production rigor. After a design is validated, established workflows for code generation, verification, and compliance are applied.
Using tools such as Embedded Coder®, teams can automatically generate optimized C and C++ code for deployment on production hardware and then apply proven verification workflows to establish confidence for deployment. AI agents can streamline this phase by automating integration tasks, building test interfaces, and orchestrating deployment workflows. Importantly, GenAI does not replace existing toolchains—it helps connect them.
Regardless of how code is created (handwritten, created by AI, or generated from models), it must be checked. Polyspace tools make it possible to analyze robustness, detect security vulnerabilities, verify compliance with coding standards, and prove absence of run-time errors. This step ensures that AI-assisted development meets the coding standards for automotive software.
Validation in a GenAI-Enabled Engineering Workflow
As the use of GenAI-created or GenAI-modified components increases, validation becomes even more critical. Automotive systems are complex and multidomain. Many defects do not appear at the level of individual functions or components. Instead, they surface during integration—when software features interact with each other, sensors, actuators, and real-world operating conditions.
This challenge is amplified with GenAI. When parts of a software feature are generated or modified automatically, reviewers must evaluate not only the change itself but also its effects on system behavior, interfaces, tests, and integration assumptions.
The validation stage is exactly where simulation is critically needed, by enabling engineers to simulate and test software together with electrical, mechanical, and environmental conditions long before physical prototypes are available. Simulation, therefore, is the mechanism that exposes integration issues early and supports “shift‑left” validation.
MathWorks is accelerating the delivery of GenAI capabilities while providing the stability and quality required for automotive engineering. Two high-quality releases a year provide a trusted foundation, while faster iteration cycles of GenAI offerings allow new GenAI capabilities to be delivered rapidly. The current GenAI capabilities include MATLAB, Simulink, and Polyspace agentic toolkits as well as Simulink Copilot.
GenAI is changing how automotive software and systems are created. Success depends on balancing innovation with rigor.
Teams that apply GenAI without enforceable verification and traceability will create the risk of more rework during integration, certification, and maintenance as well as high token usage. Teams that apply GenAI within Model-Based Design workflows can scale speed with confidence because every change is traceable and anchored in executable behavior and validated evidence.
By combining GenAI with Model-Based Design, teams can accelerate creation while preserving the executable behavior, verification evidence, traceability, and deterministic implementation needed for safety-critical engineering. MathWorks is committed to partnering with the automotive industry to build GenAI-enabled engineering environments that meet quality expectations and safety needs and ensure long-term maintainability.
