A
Jaguar Land Rover body shop in Solihull now runs a quality inspection process
in which AI-powered vision systems examine every weld on every vehicle,
autonomous agents log defects, generate repair instructions, reroute affected
vehicles on the production line, and update supplier quality databases, all
without a human reviewer seeing any individual result unless the defect
severity exceeds a threshold defined by engineers. The entire cycle takes
under four seconds per vehicle. The human inspectors who previously performed
this work have been redeployed to the exception cases the AI flags for human
judgement. This is not an experimental deployment. It is current industrial
practice, and it is representative of what is happening across advanced
manufacturing in the UK and globally as agentic AI moves from demonstration
to operational standard.
Agentic AI in manufacturing differs categorically from the AI
tools that dominate public discourse. It does not answer questions or
generate text. It plans and executes multi-step processes toward defined
goals, monitors its own performance, adapts to changing conditions, and
interacts with physical systems through robotics, conveyors, and industrial
control infrastructure. The operational environment is the factory floor, not
the browser tab, and the consequences of errors are physical and immediate
rather than informational and correctable. This raises the stakes of both the
technology and its governance considerably.
Quality Control and Inspection
Visual quality inspection is the most widely deployed agentic AI
application in manufacturing because its value proposition is unambiguous.
Machine vision systems running convolutional neural networks can examine
components, welds, surface finishes, and assembly configurations with
consistency, speed, and sensitivity that human inspectors cannot match over
sustained production shifts. A camera inspecting semiconductor wafers can
evaluate thousands of features per image at rates measured in milliseconds; a
human inspector examining the same features would need seconds, could not
maintain the same concentration across an eight-hour shift, and could not
simultaneously consult the full history of defects on that specific
production line. The business case is strong, and deployment has accelerated
accordingly. Research from the Make UK manufacturing association
found that over 60 percent of UK manufacturers with revenues above 50 million
pounds had deployed or were actively piloting AI quality inspection systems
by 2024.
What makes these systems agentic rather than simply automated is
their capacity to take consequential action on inspection findings rather
than simply logging results for human review. A defect detection system that
flags a result for a human to act on is an inspection tool. A system that
detects a defect, identifies its root cause from production data, generates a
corrective action, adjusts upstream process parameters, and communicates with
the supplier whose component caused the issue is an agentic system. The
latter is where manufacturing AI is increasingly moving, and the transition
from the former to the latter is happening faster in sectors including
automotive, aerospace, and electronics than the manufacturing policy debate
has yet absorbed.
Predictive Maintenance and Asset Management
Predictive maintenance was one of the earliest and best-evidenced
applications of machine learning in manufacturing, and agentic AI is now
extending it significantly beyond its original scope. Classic predictive
maintenance systems monitor sensor data from equipment and alert human
maintenance teams when indicators of impending failure are detected. Agentic
predictive maintenance systems go further: they identify the failure, assess
its severity and timeline, check parts inventory, generate a work order,
schedule the maintenance window to minimise production disruption, order
replacement parts if inventory is insufficient, and brief the maintenance
team on the specific procedure required. The human team executes the
maintenance; the agentic system handles all of the surrounding logistics and
decision-making.
The productivity gains from this approach are substantial and
well-documented. Siemens, which has deployed agentic maintenance systems
across several of its own manufacturing facilities as well as selling them to
customers, reports reductions in unplanned downtime of 30 to 50 percent in
facilities where the systems have been fully integrated. Unplanned downtime
is typically the single most expensive operational disruption in
manufacturing, and reductions of this magnitude translate directly into
significant profitability improvements. The return on investment case for
agentic maintenance systems in high-value manufacturing environments is
generally compelling within two to three years of deployment, which explains
why adoption rates are rising rapidly despite the significant integration
complexity involved.
Supply Chain and Production Planning
Agentic AI is also being deployed in manufacturing supply chain
management and production planning with consequences that extend beyond
individual factory walls. AI agents that monitor supply chain conditions,
including supplier capacity, logistics delays, commodity price movements, and
demand signals, can adjust production schedules, renegotiate delivery windows
with customers, and resequence manufacturing priorities in ways that maintain
output targets under conditions that would previously have required days of
manual replanning. During the supply chain disruptions of the post-pandemic
period, manufacturers with agentic planning systems demonstrated
significantly greater resilience than those relying on traditional planning
software and manual decision-making. The competitive advantage this provides
is increasingly being recognised as a structural differentiator between
manufacturers who can absorb supply disruptions and those who
cannot.
The labour implications of agentic manufacturing AI are real and
require honest acknowledgment alongside the productivity narrative. The roles
most directly displaced are those involving routine monitoring, inspection,
and logistics coordination that agentic systems now perform autonomously.
Make UK’s workforce survey data suggests that manufacturing AI is
simultaneously creating demand for higher-skill roles in AI system
management, data science, and advanced engineering while reducing demand for
the entry-level and semi-skilled roles that have historically provided
pathways into manufacturing employment. Managing this transition equitably,
through investment in workforce retraining and skills development that the
productivity gains from AI could fund, is a policy and corporate
responsibility challenge that the manufacturing sector has not yet addressed
at adequate scale. The workforce transition challenge in manufacturing
deserves more direct policy attention than it currently receives. The
Automotive Transformation Fund and the Made Smarter programme both provide
some support for AI adoption in UK manufacturing, but neither is primarily
designed to manage the workforce consequences of that adoption. A dedicated
manufacturing AI transition fund that links productivity support to workforce
investment requirements, conditioning access to technology adoption grants on
commitments to retrain displaced workers at living wage replacement rates,
would better align the incentives of employers, workers, and government.
Several European countries including Germany and Denmark have implemented
similar conditionality frameworks in their industrial AI support programmes,
with early evidence that the approach produces better workforce outcomes
without significantly reducing adoption rates. For related analysis, see our
coverage of agentic
AI and edge computing and the
automation divide.
What This Means for You
The goods you buy are increasingly manufactured with significant
agentic AI involvement in their production process. The reliability
improvements and cost reductions that agentic quality control and maintenance
systems deliver show up, eventually, in product quality and price. The jobs
displaced by agentic manufacturing systems affect communities with
concentrations of manufacturing employment in ways that are not adequately
captured by aggregate productivity statistics. Both consequences are real,
and both deserve attention from policymakers who are currently more focused
on the productivity opportunity than the workforce transition challenge. The
manufacturing sector’s adoption of agentic AI is proceeding at a pace that
policy is not keeping up with, and the governance frameworks needed to ensure
that the transition is managed equitably have not yet been developed with the
urgency the situation requires.
The Make UK manufacturing
body has called for a dedicated industrial AI transition fund that links
productivity grants to workforce investment commitments, a policy proposal
that has cross-party support in principle but has not yet been implemented at
adequate scale. The competitive pressure on UK manufacturers from AI-enabled
rivals in Germany, the Netherlands, and increasingly from advanced
manufacturing facilities in Asia means the window for managed transition is
shorter than the policy timeline suggests.
About the Author
Stuart Kerr is a technology correspondent at LiveAIWire, covering
artificial intelligence, digital innovation, and the social impact of
emerging technologies. Follow LiveAIWire for daily analysis at liveaiwire.com.