Artificial intelligence is entering a new phase in computer-aided design and manufacturing (CAD/CAM). The early focus on generative design—using AI to produce shapes and configurations from text prompts—is giving way to something more consequential: AI agents that operate engineering software, run analyses, develop manufacturing plans, and connect design decisions to physical production.
Developments from Autodesk, Siemens, PTC, CloudNC, and emerging engineering-AI companies illustrate this shift. Shipbuilders including HII, Fincantieri, HD Hyundai, Hanwha Ocean, and Anduril are also pursuing automated production systems that connect digital engineering with robotics, inspection, and manufacturing execution.
Three trends stand out:
- AI is moving inside engineering applications, using native CAD, CAM, and simulation tools rather than generating isolated design artifacts.
- The digital thread is becoming executable, connecting requirements, geometry, analysis, manufacturing instructions, inspection, and configuration records.
- Physical AI is reaching the factory floor, where adaptive robots and intelligent process controls can respond to actual production conditions rather than following fixed instructions.
Together, these developments have significant implications for naval engineers, manufacturing practitioners, and engineering educators.
AI Becomes an Engineering Agent
Autodesk's September 2026 announcements show how quickly engineering software is evolving. Emerging Fusion capabilities include AutoTimeline, which reconstructs editable parametric histories from imported geometry; AutoAssemble, which identifies assembly relationships; and AI-assisted CAM functions that generate machining strategies and toolpaths.
PTC's Onshape takes a complementary approach. Its FeatureScript MCP Server allows AI agents to create reusable parametric design functions, encoding recurring engineering practices as tools for other designers.
In October, MecAgent demonstrated an AI agent working within SolidWorks to build and modify components and assemblies. A roughly 56-component helicopter model illustrated the potential for sustained CAD automation, though it was not a validated, flight-capable design.
These developments signal a shift in engineering practice: AI can increasingly execute modeling sequences, inspect results, correct errors, and preserve editable engineering artifacts. For naval architects and marine engineers, the near-term opportunity is not autonomous ship design but automation of repetitive work such as machinery foundations, pipe supports, structural details, equipment arrangements, and design modifications.
AI Connects CAD to Analysis and Manufacturing
The next step is to integrate engineering agents with established analysis and manufacturing tools. Ansys has introduced an interface that lets AI agents operate its Mechanical APDL structural-analysis environment. SimScale is similarly integrating AI with cloud-based computational fluid dynamics and finite-element analysis. In both cases, agents can prepare models, run established solvers, interpret results, and support design iteration.
The distinction matters: AI manages the workflow, while validated engineering software performs the calculations. Manufacturing applications are advancing as well. CloudNC's CAM Assist generates machining strategies and toolpaths; its October Quote Agent release extends AI into manufacturability assessment, process planning, cost estimation, and inspection preparation.
Siemens' NX Manufacturing developments likewise incorporate AI-assisted machining recommendations, tool selection, and process planning. Its digital-machine initiatives allow manufacturers to simulate and validate processes before installing equipment.
The resulting workflow is increasingly continuous:
Engineering requirements → CAD model → analysis → manufacturing plan → CAM → machine simulation → production → inspection.
The potential gain is not merely faster modeling. A connected workflow can reduce delays and information loss between engineering disciplines and production organizations.
Shipbuilding Moves Toward Physical AI
Shipbuilding is a demanding test because its production environment differs fundamentally from repetitive automobile manufacturing. Ships are large and highly variable; structures distort, access is restricted, and actual conditions routinely diverge from nominal CAD geometry.
Even so, significant initiatives are underway. HII's HYPR initiative, involving Path Robotics and GrayMatter Robotics, establishes performance-based agreements worth up to $900 million for qualified robotic production work. HII is also testing intelligent welding technologies with HD Hyundai at Ingalls Shipbuilding. Hanwha Ocean reports that AI assists approximately 67% of indoor welding activities at its Geoje shipyard, though that figure does not mean the welding is fully autonomous. Fincantieri's October BOOST² initiative is expanding physical AI, robotics, digital twins, and advanced manufacturing across 13 shipyards, including U.S. facilities, with nearly $100 million in planned investment through 2030.
New entrants are taking different approaches. Anduril's proposed Arsenal-2 submarine facility would use an integrated, software-defined manufacturing system, while Saronic's Port Alpha project envisions a digitally enabled greenfield shipyard.
These efforts share an objective: connect the authoritative product model to adaptive production equipment and automated quality evidence. The question is no longer whether robots can weld, grind, or inspect. It is whether they can do so economically across changing ship configurations while meeting qualified production standards.
Digital Twins Become Engineering Decision Systems
Digital twins are likewise evolving beyond three-dimensional visualization. Navantia has reported substantial reductions in selected engineering cycle times using an integrated Siemens Teamcenter environment that connects requirements, systems engineering, CAD, simulation, and lifecycle data. In September, ABS, HD Korea Shipbuilding & Offshore Engineering, and Siemens announced work combining computational fluid dynamics and digital twins to assess ammonia leakage and shipboard safety.
Other efforts connect laser scanning and reality capture with authoritative engineering models, allowing organizations to compare the as-designed and as-built states. In shipbuilding, these capabilities could improve block alignment, outfitting verification, production sequencing, engineering-change assessment, and maintenance planning. The larger opportunity is a decision-supporting digital twin that identifies deviations, evaluates alternatives, predicts consequences, and records verified outcomes.
What the Research Tells Us
Academic research offers both promise and caution. Recent work on DeciCAD, TraceCAD, and other engineering-specific agents shows that AI performs better when it uses structured CAD commands, preserves modeling history, executes models, and corrects errors. Research on automated assembly planning also shows that systems can derive production sequences, tool requirements, and work instructions from product geometry.
Yet benchmarks such as CADWorld reveal substantial limitations. In September testing, the strongest evaluated general-purpose agents completed only 17.5% of realistic engineering tasks, compared with 87% for human experts. A visually plausible model is not necessarily a valid engineering design. Dimensions, tolerances, materials, loads, interfaces, manufacturability, and configuration control still matter.
AI-generated engineering must be verified against engineering requirements—not judged by the appearance of its output.
Engineering Impacts: Practice, Education, and Professional Responsibility
For practicing engineers, these developments suggest three changes.
- First, engineers will increasingly supervise intelligent tools that handle routine modeling, analysis preparation, manufacturing planning, and documentation. Their work will shift toward defining requirements, evaluating alternatives, verifying results, and resolving exceptions.
- Second, digital engineering competence will become inseparable from traditional engineering knowledge. Engineers will need to understand authoritative data, model-based definition, configuration management, simulation, software interfaces, and the connections between design and production systems.
- Third, the engineer's responsibility for technical correctness remains unchanged. AI may accelerate an analysis or generate a manufacturing plan, but it cannot establish that the result is safe, producible, or fit for purpose.
Engineering curricula should evolve accordingly. Students should learn to use AI-enabled CAD/CAM and simulation tools while demonstrating independent competence in mechanics, materials, manufacturing processes, systems engineering, and verification.
One useful exercise would compare a manually developed design with an AI-assisted alternative—not only on completion time, but also on geometric validity, analytical assumptions, manufacturing feasibility, configuration traceability, and engineering judgment.
The goal is not to produce engineers who merely know how to prompt AI. It is to produce engineers who understand how to direct, evaluate, and verify increasingly automated engineering systems.
The Larger Transformation
The Fourth Industrial Revolution in maritime engineering is more than another generation of design software. It is beginning to connect engineering knowledge, computational analysis, manufacturing equipment, robotics, inspection, and operational feedback into a continuously improving industrial enterprise. For naval engineering, the decisive measure will not be how quickly AI produces a drawing or how many robots a shipyard installs.
It will be whether these technologies help the maritime industrial base design, manufacture, verify, deliver, and sustain more capable ships and systems faster, with greater quality and reliability. That is the promise of AI-enabled CAD/CAM—and the challenge now facing engineering practitioners and educators.
Main image caption: Vincent Malpaya, mechanical engineer with the Airborne Telemetry Group, compares a 3D-printed prototype to its computer-aided design (CAD) model at the Innovation Lab. (U.S. Navy photo by Ryan Figueroa)