digital.auto
Agentic Engineering
A PARTNERSHIP FOR THE FUTURE OF
VEHICLE DEVELOPMENT
WORKSTREAM
digital.auto and the Agentic Engineering initiative are joining forces to bring
Agentic Engineering into automotive practice.
The partnership connects methods and research across engineering domains with digital.auto’s automotive ecosystem and hands-on experimentation.
Agentic Engineering brings engineers, specialized AI agents and executable models together across requirements, architecture, implementation, integration and validation. Agents help interpret engineering intent, develop solutions and test their consequences—with human oversight and clear responsibilities.
Our shared ambition: make vehicle development more connected, accelerate feedback and uncover integration problems earlier. Through digital.auto’s Agentic Engineering workstream, we are translating this ambition into practical methods, shared demonstrators and reusable engineering patterns.
Loop Engineering: Turning AI Capabilities into Engineering Outcomes
At the heart of Agentic Engineering is the Engineering Loop:
Understand → Generate → Integrate → Validate → Learn → Improve.
Each iteration connects a proposed change with evidence of how it affects the system, feeding the results back into the next engineering decision.
Loop Engineering is the discipline of designing and governing these loops: defining objectives, connecting agents to tools and knowledge, selecting validation methods, and setting approval boundaries and stopping criteria. Success depends on the quality of the complete loop—including its feedback.
In vehicle development, specialized loops can address functions, software, calibration or testing. Stable interfaces and abstraction layers allow these loops to progress at different speeds while remaining connected to shared vehicle requirements and system-level validation.
REASONING SURFACES
AN EXECUTABLE VEHICLE FROM THE START
Engineering agents need a shared environment in which to understand dependencies, explore alternatives and evaluate changes. We call this a Reasoning Surface: connected engineering knowledge combined with executable models and feedback from tests and simulations.
For automotive engineering, this requires an early-stage, lightweight 360° vehicle scaffold, available already during requirements and architecture work on the left side of the V-Model. It provides broad coverage of the vehicle’s functions, interfaces and interactions, initially using simplified models and explicit assumptions. Engineers and agents can exercise end-to-end scenarios and investigate integration issues before detailed implementations or physical prototypes exist.
Building on SDV Forge, Agentic Forge develops this approach toward a progressively richer reasoning surface. Higher-fidelity simulations, real software and hardware replace simplified elements as development advances. Early findings remain tied to the capabilities and limitations of the models that produced them.
Agentic Forge Whitepaper Series
From Integration Hell to continuous, agentic engineering.
Agentic Forge is the central initiative within the Agentic Engineering Work Stream.
It evolves the work started with SDV Forge: bringing integration forward and creating an executable representation of the complete vehicle before the physical vehicle exists. The initial concept uses engineering artifacts such as requirements, function descriptions and interfaces as input for creating that early model.
Agentic Forge takes the next step.
AI agents can help turn fragmented engineering information into executable artifacts, reason across dependencies and domains, support integration and validation, and continuously incorporate what is learned back into the engineering process.
The vehicle becomes executable earlier.
Integration becomes continuous.
AI becomes part of the engineering system.
The series began as SDV Forge, focusing on the industry's integration challenge and the need to move integration and validation earlier into the development lifecycle.
With Parts 2-4, the perspective expands toward Agentic Engineering and the AI-defined vehicle.
00
Executive Summary
The starting point: why traditional engineering approaches struggle with the complexity and integration demands of the Software-Defined Vehicle.
01
From Integration Hell to Continuous Integration
Why late integration has become a critical bottleneck - and how an executable vehicle can shift integration and validation left.
02
Coming Soon
Stay Tuned
Agentic V-Loops
AGENTIC V-LOOPS COMBINE THE STRUCTURE OF THE TRADITIONAL V-MODEL WITH ITERATIVE FEEDBACK LOOPS FOR AGENTIC ENGINEERING.
Requirements, decomposition, integration and verification remain essential. Agentic loops connect these activities continuously, allowing findings from executable models and tests to inform requirements, architecture and implementation much earlier.
A proposed function can be generated, integrated into the vehicle scaffold, evaluated against defined scenarios and refined through successive iterations. As the system matures, the same approach extends to software and hardware integration, regression testing and the evidence needed for homologation—with traceability, review gates and human accountability.
The digital.auto workstream brings together OEMs, suppliers, tool providers and researchers to put these concepts into practice through demonstrators and the Agentic Forge whitepaper series.
Meet the Agentic Engineering Workstream Chair
DRIVING THE INITIATIVE
Chair
Dirk Slama
Work Stream Lead, Agentic Engineering
Chairman, digital.auto
Dirk Slama leads the digital.auto Agentic Engineering Work Stream and the evolution of The Forge.
The Work Stream builds on his work around the SDV integration challenge and expands that thinking toward agentic engineering and the AI-defined vehicle.
Agentic Engineering Workstream participants
