By VADM David Lewis (USN, Ret.), President of the American Society of Naval Engineers
Top 3 takeaways
- AI is becoming an execution layer. The most useful tools are connecting to project data, design environments, CAM workflows, CNC verification, and operational feedback.
- Digital-thread readiness matters more than model novelty. Governed data, permissions, simulation confidence, and configuration control are now prerequisites for safe AI deployment.
- Manufacturing assurance is a near-term win. AI-assisted CAM, machine twins, post-processing validation, and CNC verification can reduce risk before autonomous design becomes production-ready.
AI implementation is becoming a digital-thread problem, not a CAD feature problem. The strongest recent developments point toward connected, governed execution layers: AI agents tied to CAD/CAM environments, PLM data, industrial data models, CNC verification, digital twins, asset performance, and production operations.
Bottom line: The winners will be the engineers and organizations that prepare the data layer, simulation layer, workflow governance, and production feedback loop before expecting AI agents to act safely at scale.
“AI value in engineering will come less from isolated demos and more from trusted, connected workflows that can survive real production constraints.”
Autodesk: AI moves from CAD action into project-data control
Autodesk introduced the Fusion Data MCP Server on 2 July 2026, a cloud-hosted Model Context Protocol server that lets AI clients access Fusion backend data capabilities such as hubs, projects, folders, items, permissions, and team-member access. This complements the earlier Fusion MCP, which connects AI tools to a live Fusion session for geometry, modeling operations, and design-state interaction.
Why it matters: This is a practical step toward AI as a CAD/CAM workflow operator rather than a mere assistant. In ship design, aerospace structures, or unmanned-system development, the bottleneck is often not just making geometry; it is controlling project data, permissions, file structures, handoffs, and configuration discipline.
Siemens + Microsoft: Cloud-based AI design-to-CAM on Azure
Siemens published a 1 July 2026 update on a Siemens–Microsoft Hannover Messe demonstration using NX X, the cloud SaaS version of NX, including both design and CAM, running on Microsoft Azure. The workflow showed a humanoid-robotics component moving from plain-language requirements to design options, then to AI-generated machining strategy, tooling recommendations, and machining parameters. Siemens also emphasized that the system can learn from previous jobs and adapt to a shop’s standards and workflows.
Why it matters: this is close to the operating model defense manufacturing needs: requirements-to-geometry-to-toolpath inside a governed cloud engineering stack. The immediate relevance is not “AI designs a ship,” but “AI reduces the time from design intent to machine-ready work package” for brackets, fixtures, robotics components, drone parts, shipyard tooling, and precision-machined assemblies.
Siemens + IFS: Closing the loop between design, production, and field performance
Siemens and IFS announced a strategic partnership to connect engineering intelligence with operational reality across the product lifecycle. Siemens brings industrial AI, engineering, automation, manufacturing execution, and digital-twin context; IFS brings enterprise asset management, field service, asset behavior, service history, and operational lifecycle data. Their stated goal is a closed-loop digital twin grounded in both design intent and field performance.
Why it matters: this is a major “digital thread after delivery” development. For naval ships, aircraft, ground vehicles, and industrial equipment, the most valuable AI may be the one that learns from operations, maintenance, failures, inspections, and service history, then feeds those lessons back into design, production planning, and sustainment. This is directly relevant to fleet maintenance, shipyard throughput, and reliability-centered design.
Schneider Electric + Cognite: Industrial AI consolidation accelerates
Schneider Electric announced that it agreed to acquire Cognite for $3.1 billion. Cognite’s platform provides industrial data contextualization, a unified industrial data model, knowledge graph capabilities, and agentic AI through Atlas AI; Schneider plans to integrate Cognite with AVEVA’s CONNECT industrial intelligence platform across design, build, operation, and optimization.
Why it matters: this is not CAD/CAM software narrowly defined, but it is very important for digital engineering implementation. Industrial AI needs contextualized plant, factory, energy, equipment, and process data. Cognite gives Schneider/AVEVA a stronger foundation for AI that can operate across engineering data, operational data, asset data, and plant-floor workflows. For shipyards and defense industrial facilities, the lesson is clear: AI value depends on the industrial data foundation, not just the model.
Nidec: Digital twins move into large-machine CNC prove-out
Nidec Machine Tool launched Nidec NC Twin on 1 July 2026 for its MVR Series double-column five-face milling machines. The platform recreates machine movement and processes in a virtual environment, allowing users to validate NC programs before production, check interference, simulate machining time, and predict surface quality. Nidec says it uses the same NC program as the actual machine and, for machining-time estimation, can keep cycle-time error within 1% excluding auxiliary-equipment motion errors.
Why it matters: large ship, submarine, aerospace, and energy components often suffer from long prove-out cycles, expensive machine downtime, and limited expert machinist availability. Nidec NC Twin points toward a practical AI-adjacent pattern: build high-fidelity machine twins first, then layer AI onto verified simulation, cycle-time prediction, fixture planning, and process optimization.
Vericut: AI assistance expands in CNC verification and post-processing
Vericut 9.7 was announced in late June with expanded AI-powered assistance, faster CNC simulation and verification workflows, and context-aware Vericut Assistant guidance inside the software. The associated Icam Post V27 release adds AI-powered help for post developers, including product guidance, macro syntax support, regression testing, and validation workflows for post-processor changes.
Why it matters: this is the less glamorous but highly valuable side of AI-enabled CAM. The defense industrial base does not merely need faster toolpath generation; it needs safer NC verification, fewer collisions, better G-code confidence, and controlled post-processor changes. Vericut’s direction supports AI as a trusted manufacturing assurance layer.
Dassault Systèmes: Virtual-twin foundations for complex infrastructure
Dassault Systèmes and UK Fusion Energy Ltd announced next steps to advance the UK’s STEP prototype fusion plant using the 3DEXPERIENCE platform. Dassault described the effort as building a virtual-twin and digital-engineering foundation that keeps plant information, engineering decisions, and system designs connected across the lifecycle.
Why it matters: this is a useful analog for large naval and industrial programs. Fusion plants, nuclear ships, submarines, carriers, and complex yards all require lifecycle digital continuity across systems engineering, plant design, manufacturing, supply chain, configuration control, and operations. The relevance is not the fusion plant itself; it is the use of virtual twins to industrialize highly complex systems.
Research reality check: AI-CAD is improving, but production CAD still needs validation
Recent AI-CAD research continues to show progress but also reinforces that industrial CAD is hard because it requires editable, parametric, feature-based, manufacturable models rather than visually plausible 3D shapes. A June 2026 AI+CAD paper argues that CAD data representation is more foundational than model optimization for industrial-grade parametric feature modeling. IterCAD, another June 2026 paper, advances closed-loop, multimodal CAD generation and editing by combining an executable CAD sandbox with iterative refinement and geometry-aware reinforcement learning.
Why it matters: this supports a disciplined adoption strategy. Use AI now for design review, DFM checks, project-data management, CAM suggestions, CNC verification, simulation acceleration, and digital-thread navigation. Treat fully autonomous text-to-production CAD as an R&D frontier, not a production assumption.
Main image caption: Aviation Electricians Mate 1st Class Corie Wooldridge, assigned Wasp-class amphibious assault ship USS Boxer (LHD 4), gives training on 3D printing software to U.S. Marine Corps Cpl. Simon Quilacio, an avionics electrician assigned to Marine Medium Tiltrotor Squadron (VMM) 163 (Reinforced), in the ship’s calibration lab, April 2, 2026. Boxer, flagship of the Boxer Amphibious Ready Group, is underway with the 11th Marine Expeditionary Unit in the U.S. 3rd Fleet area of operations demonstrating the U.S. Navy’s long-term commitment to a free and open Indo-Pacific. (U.S. Navy photo by Mass Communication Specialist Seaman Dustin Drake)
What this means for maritime, aerospace, and defense
The emerging implementation pattern: requirements and intent → CAD/CAE/MBSE → CAM/toolpaths → CNC/digital twin prove-out → production execution → inspection/as-built data → sustainment feedback → design update.
For naval applications, the priority should be to connect AI to governed data and validated workflows: ship-design repositories, PLM, MBSE models, work packages, yard schedules, machine twins, robotic work cells, inspection records, and maintenance history. The near-term payoff is faster design iteration, fewer late manufacturability surprises, shorter CNC prove-out, better configuration control, and tighter feedback from fleet operations into future design.
Closing takeaway: AI will matter most where it is connected to trusted engineering data, validated simulations, controlled manufacturing workflows, and feedback from real-world operations.
For readers
If you are evaluating AI for engineering or manufacturing, start with the workflow you need to govern: data access, model validation, simulation fidelity, shop-floor handoff, inspection feedback, and sustainment learning. The technology is advancing quickly, but adoption will depend on whether organizations can make AI traceable, auditable, and useful inside real engineering constraints