Abstract
As Automotive SPICE (ASPICE) requirements have evolved with PAM 4.0 and engineering systems continue to grow incomplexity, organizations are increasingly adopting AI-augmented process reviews to improve quality, speed, and consistency.
Using Large Language Models (LLMs), the solution enables advanced capabilities such as image-based architecture analysis, workflow extraction and comparison, structured table interpretation, and cross-document reasoning.
Building on more than three years of real-world deployment experience, this presentation demonstrates how AI-enabled work product reviews have evolved to a production-ready solution with multiple global embedded systems programs.
The tool was enhanced in the latter part of last year to support User feedback, Model Training, and Expert-in-loop reviews through workflows.
The session will present the transition from ASPICE PAM 3.1 to PAM 4.0 on global embedded projects. Particular focus will be placed on how a continuous feedback loop—leveraging real-time assessment experiences, assessor insights, and process-specific training—helps improve the AI agent’s accuracy, relevance, and adaptability while aligning with the evolving expectations of PAM 4.0 assessments.
Key Takeaways
- Real deployment insights on improving review speed, consistency, and assessment readiness.
- How AI–human collaboration reduces manual workload without compromising engineering.
- How the AI work product reviewer is being upgraded to ASPICE PAM 4.0.
- How real assessment feedback and assessor expertise continuously train the AI agent to improve accuracy
As Functional Safety (ISO 26262) requirements are highly critical and embedded automotive systems continue to grow in complexity, organizations are increasingly adopting AI augmented safety process and work product reviews to improve quality, speed, and consistency.
Leveraging Large Language Models (LLMs), the solution enables advanced capabilities such as image based architecture analysis, safety concept interpretation, workflow extraction and comparison, structured table analysis, and cross document reasoning across safety artifacts.
Building on more than three years of real world deployment experience with an AI based ASPICE Work Product Reviewer, this presentation demonstrates how AI can assist preliminary Functional Safety verification and confirmation reviews of ISO 26262 work products.
At the beginning of this year, the first ISO 26262 Part 4, Part 5, and Part 6 use cases were implemented and successfully tested in initial projects.
The session will illustrate how AI can effectively support the consistent application of ISO 26262, strengthen safety argumentation, and scale Functional Safety activities across global development organizations.
Value
Organizations can learn how to leverage new technologies such as AI to apply it into their business workflows and optimize existing work. The session will also emphasize use of such technology for preventive quality assurance that empowers development teams during new product development.
The presentation will also highlight how real assessment feedback and assessor expertise is used to continuously train the AI agent to improve its accuracy.
Key Takeaways
- How AI–human collaboration reduces manual workload without compromising engineering.
- Practical integration of ASPICE and ISO 26262 work product reviews using a unified AI approach
- How can the AI Agent further assist a comprehensive preliminary Functional Safety verification review and confirmation review of Functional Safety work products