In the fast-evolving world of aircraft manufacturing, precision, efficiency, and safety remain paramount. Today’s wave is not just Machine Learning (ML) – it is foundation-model copilots embedded in engineering tools, Open Universal Scene Description (OpenUSD)-based digital twins that tie the digital thread end-to-end, and edge robotics artificial intelligence (AI) that brings real-time intelligence to the factory floor.
Copilots inside Product Lifecycle Management (PLM), Computer-Aided Design (CAD) & Computer-Aided Manufacturing (CAM)
Modern platforms now ship with AI assistants that understand your product data and process context:
Teamcenter Copilot helps engineers query PLM data and standards in natural language and automate routine tasks across the lifecycle.
NX Design/Manufacturing Copilot brings AI help directly into CAD and (soon) CAM programming, speeding modeling and toolpath setup while reducing rework.
Onshape AI Advisor targets real-time design guidance in the browser; useful for agile, multi-party collaboration.
Autodesk Fusion AI advances generative design, simulation guidance, and automation across “Design & Make” workflows.
What this changes: Designers can get Design for Manufacturability/Design for Assembly (DFM/DFA) tips, tolerance hints, and process-aware suggestions inside the tools they already use – while referencing the company’s own standards and prior art.
Digital Reality Twins with OpenUSD
Photorealistic, physics-aware digital twins are moving from demo to daily work. Siemens is delivering a Teamcenter Digital Reality Viewer powered by NVIDIA Omniverse; the broader industry is rallying on OpenUSD for interoperable three-dimensional (3D) scenes and factory models.
What this changes: Designers and manufacturing engineers co-review the same live twin—layouts, clearances, human factors—before hardware, reducing late-stage changes and travel for line trials.
Robotics, Simulation & Edge AI
Robotics is benefiting from simulation-first workflows and new “robot brains”:
Isaac Sim generates synthetic data and validates flows in photorealistic environments; vendors use it for collaborative robots (cobots) and autonomous mobile robots (AMRs).
GR00T/robotics foundation models are emerging, pointing toward generalist manipulation; NVIDIA’s latest Jetson Thor pushes far more on-edge reasoning for real-time tasks.
What this changes: Higher-mix cells and inspection tasks get faster to deploy—train in simulation, transfer to shop, iterate with data loops.
Quality & Safety: From Paper to Data Products
Aerospace quality is formalizing the digital thread around well-known standards:
AS9145 (Advanced Product Quality Planning, APQP, with Production Part Approval Process, PPAP, adapted for aerospace) is increasingly baked into gated programs;
AS9102 First Article Inspection (FAI) flows are becoming model-based, not just forms.
Quality Information Framework (QIF, ISO 23952) is the metrology backbone linking Product Manufacturing Information (PMI)/Geometric Dimensioning & Tolerancing (GD&T) to inspection plans and results – closing the loop back to PLM.
What this changes: ML/Large Language Model (LLM) systems can auto-assemble FAI packages, balloon drawings, and inspection plans from authoritative CAD/PMI and QIF data – then reconcile measured results to tolerances, flagging drift early.
Governance & Certification Are Catching Up
If you deploy AI in safety-relevant contexts (e.g., Numerical Control (NC) programming suggestions, inspection, maintenance recommendations), expect rising governance:
The European Union Artificial Intelligence Act (EU AI Act) classifies many manufacturing/safety components as high-risk, triggering documentation, testing, and conformity-assessment duties.
European Union Aviation Safety Agency (EASA) AI roadmap and concept papers detail assurance paths for ML applications in aviation.
Systems Modeling Language version 2 (SysML v2) was approved in July 2025, bringing standardized Application Programming Interfaces (APIs) to Model-Based Systems Engineering (MBSE) and making it easier to bind requirements, hazards, tests, and work instructions into one model.
What this changes: Your AI assistants should be traceable (inputs, model versions, prompts), auditable, and linked to requirements, risks, and verification artifacts in MBSE/PLM.
What “good” looks like now (2025)
Design with guardrails: Use CAD/PLM copilots that query internal standards, approved processes, and lessons learned. Back them with Retrieval-Augmented Generation (RAG) over your PLM/wiki/Standard Operating Procedure (SOP) corpus—so advice is sourced, not speculative.
Sim-first production changes: Prove fixture reach, takt, and safety in an OpenUSD/Omniverse twin before moving a machine.
Automated quality loops: Drive inspection from PMI/QIF, generate AS9102 artifacts automatically, and close the loop by pushing deviations to Corrective and Preventive Actions (CAPA) and design updates.
Edge intelligence: Where milliseconds matter (vision inspection, cobot hand-offs), deploy models on Jetson Thor/Orin-class edge computers and keep the twin in sync.
Compliance by design: Treat the EU AI Act + EASA guidance as non-functional requirements. Log datasets, approvals, and substantial-change events like you do for any regulated part.
A Safer, Faster “Artistry + Precision” Workflow
The promise holds: designers create boldly with manufacturability insight; technicians execute precisely with design intent visible; and safety is strengthened by data-centric quality and clear AI traceability. With copilots in your engineering stack, OpenUSD twins for shared truth, and edge AI on the line, the gap between concept and certified part keeps shrinking – without compromising the rigor aerospace demands.
Conclusion
This is no longer a “wait-and-see” moment – it is a build-and-measure moment. By putting artificial intelligence (AI) copilots inside everyday engineering tools, using OpenUSD-driven digital twins as a shared source of truth, and deploying edge intelligence on the line, you turn standards into software, tribal knowledge into data, and intent into audited outcomes.
The result is fewer loops between concept and certified part, higher first-pass yield, and safety that’s demonstrated – not just declared – through traceable, model-based quality. Organizations that move now will design bolder without losing manufacturability, execute faster without sacrificing rigor, and meet emerging compliance with confidence. Choose one high-impact flow – like toolpath programming, automated inspection, or changeover validation – wire it to your digital thread, and let the assistant work. The factories that win will be the ones that make creativity computable and precision automatic.