AI Is Expanding the Definition of Industrial Automation
Artificial intelligence is no longer confined to automotive assembly lines, semiconductor fabs, or highly standardized aerospace manufacturing. The latest developments show a broader shift: automation technologies are being adapted to industries with radically different production scales, materials, processes, and operating constraints.
Two examples illustrate this change particularly well. Huntington Ingalls Industries is exploring physical AI and robotics for the production of large warship components, while Siemens is applying agentic AI to highly flexible food manufacturing, including filled cookies and donuts.
Although these applications appear unrelated, they share the same engineering objective: making automation more adaptable when conventional fixed automation becomes difficult to justify.
Huntington Ingalls Targets Large-Scale Ship Component Automation
Huntington Ingalls Industries, a major U.S. Navy shipbuilder, has entered an agreement with robotics partners to improve the fabrication of large ship components.
The proposed approach focuses on producing major assemblies away from the traditional shipyard environment before transporting them for final integration. This could change how large naval structures are fabricated by moving selected manufacturing operations into more controlled and automation-friendly environments.
From an automation engineering perspective, this is significant because shipbuilding presents very different challenges from automotive production. Components are large, geometries vary considerably, production volumes are relatively low, and processes frequently require human judgment.
Physical AI and robotics therefore need to operate beyond simple repeatable motion. Systems must interpret physical conditions, adapt to variations, and coordinate with human workers and conventional production equipment.
Physical AI Could Change Low-Volume Manufacturing
Traditional industrial robots perform best when the production environment is predictable. Shipbuilding challenges that assumption.
Large structures can have dimensional variation, irregular surfaces, changing work positions, and complex assembly sequences. A robotic system designed around fixed coordinates and rigid process assumptions can become inefficient when these variables change.
Physical AI introduces another layer of intelligence. Instead of simply executing a predetermined sequence, an automation system can potentially combine machine perception, sensor data, process models, and adaptive control to respond to changing physical conditions.
The practical value will depend less on the AI terminology and more on measurable production improvements. Position accuracy, cycle time, rework rate, operator safety, equipment utilization, and integration with existing manufacturing systems will remain the real performance indicators.
Siemens Takes AI into High-Mix Food Production
At the opposite end of the manufacturing spectrum, Siemens is applying agentic AI to food processing.
The company has collaborated with two food-industry automation firms to develop a flexible automation solution aimed at products such as filled cookies and donuts. The target environment is characterized by high product variety and comparatively low production volumes.
Food manufacturing presents a different automation problem from shipbuilding. Instead of dealing with massive structures, engineers must manage product recipes, frequent changeovers, varying ingredients, packaging requirements, sanitation procedures, and strict process consistency.
In this environment, flexibility can be more valuable than maximum machine speed.
Agentic AI Could Reduce the Cost of Changeovers
High-mix, low-volume production has traditionally created an economic challenge for automation. Dedicated machines can deliver excellent throughput, but their economics deteriorate when products change frequently.
Agentic AI could provide a different operating model by helping automation systems coordinate production decisions, equipment settings, recipes, and process sequences according to changing requirements.
The important distinction is that AI does not replace the underlying automation architecture. PLCs, motion controllers, drives, sensors, safety systems, industrial networks, and HMI platforms still provide the deterministic control layer.
AI instead operates increasingly as an intelligence layer above this infrastructure.
Two Industries, One Automation Strategy
Shipbuilding and food production appear to have almost nothing in common, yet their automation challenges reveal the same industry trend.
Both applications require machines to cope with variability. Both can benefit from combining conventional automation with perception, data processing, robotics, and AI. More importantly, neither application can depend entirely on rigid programming if manufacturers want to increase flexibility.
This suggests that the next stage of industrial automation will not simply involve replacing more manual operations with robots. The larger opportunity is creating systems capable of adapting their behavior to changing production conditions.
The Real Test Is Integration, Not AI
From an engineering standpoint, the most important question is not whether a system uses physical AI or agentic AI. The question is whether that intelligence can be integrated safely and predictably with the existing control architecture.
Industrial AI must coexist with deterministic control, functional safety, cybersecurity requirements, production databases, quality systems, and human operators. Any AI-driven recommendation that reaches the physical process must also operate within clearly defined limits.
This is particularly important in applications such as naval manufacturing, where production errors can have substantial consequences, and food processing, where product quality and process compliance are tightly controlled.
Automation Is Becoming More Adaptive
The examples from Huntington Ingalls Industries and Siemens demonstrate how automation is moving into environments previously considered difficult to automate economically.
The common direction is clear: industrial automation is evolving from fixed sequences toward adaptive production systems.
However, AI should be viewed as an additional engineering capability rather than a replacement for proven automation fundamentals. Sensors still need accurate signals, networks still need predictable communications, control systems still need deterministic behavior, and machines still require carefully engineered safety functions.
The companies that gain the most from industrial AI will likely be those that combine these fundamentals with intelligent software rather than treating AI as a standalone technology.
