Fabio Caversan is the Global Chief Technology Officer at Stefanini, driving new product offerings and digital transformation.
getty
Walk the floor of almost any major manufacturing facility, and you will find a contradiction hiding in plain view. The products coming off the line represent the latest in engineering and materials science, yet the control systems governing those lines were, in many cases, installed before the first iPhone shipped. They run on operating systems that have since been abandoned, programmed in outdated languages and are held together by spare parts that are increasingly difficult to find.
Between 50% and 70% of industrial control systems worldwide have crossed or are approaching the 20-year mark. Unplanned downtime now costs manufacturers an average of $260,000 per hour, with capital-intensive sectors absorbing losses well into the millions.
Legacy platforms were not designed in line with modern cybersecurity protections, leaving them vulnerable to intrusions like ransomware. Because these older architectures cannot interface with IIoT infrastructure or advanced analytics tools, manufacturers may also not be able to convert plant data into operational intelligence.
The traditional path for upgrading a legacy control system has always been manual, slow and expensive. Engineers study the existing code running a PLC or SCADA system, interpret what each instruction does and rewrite the entire program for the target platform. A large facility with hundreds of control points can burn through months of engineering time on conversion alone, and the risk of introducing errors during manual translation adds cost and delay at every stage.
AI tools built specifically for operational technology have reshaped that process. Current-generation platforms can ingest legacy PLC code, translate it into a vendor-neutral intermediate format, audit it for logic errors and generate equivalent code for the new system, while simultaneously producing documentation, network diagrams and I/O lists that previously required their own dedicated engineering workstreams.
Some of these platforms also include code explainers that describe legacy logic in plain language, code auditors that flag inconsistencies and PLC migrators that handle translation across vendor architectures. In large-scale modernization projects, that approach has cut timelines by half or more compared to conventional methods, with the gains scaling proportionally as plant size and complexity increase.
The efficiency improvements also extend across the broader engineering lifecycle, with leading implementations delivering significant gains across design, coding, testing and documentation workflows. Tasks that once required separate, time-intensive workstreams can now be executed in parallel or automated entirely, reducing both effort and the likelihood of human error. Engineers are able to move from initial design through validation and deployment with far greater speed and consistency, particularly in large-scale environments where complexity compounds quickly. Machine learning models are also being applied to accelerate post-migration ramp-up, compressing the time required to bring a newly modernized facility back to full production capacity.
Vendor flexibility is a big piece of this equation that rarely gets enough attention. Tools that work across platforms allow companies to evaluate which technology best fits each production environment based on cost, capability and process requirements rather than on which integrator has the preferred vendor relationship.
A production stoppage on a modern manufacturing line initiates a costly chain of dependencies. The operator registers the event, reaches maintenance and then waits for them to arrive. The technician then spends a disproportionate share of total response time on diagnostics, before any corrective action can even be scoped.
Virtual assistants engineered for shop floor deployment are restructuring that sequence by feeding operators continuous, real-time telemetry on system status and surfacing probable root causes the moment a deviation registers. The physical repair still requires a technician’s hands, but these tools reduce the investigative work that previously contributed to hours of downtime.
Aggregated across every unplanned stoppage on every line over a fiscal quarter, the recaptured production time translates into output gains that register clearly on an earnings statement.
Overall equipment effectiveness (OEE) has occupied the center of manufacturing performance measurement for more than four decades. It distills availability, throughout and product quality into a single indicator of how efficiently a plant or piece of equipment is performing.
Its longevity speaks to its usefulness, which makes it all the more notable that a substantial portion of the sector still fails to track OEE with any rigor. Without that discipline in place, every modernization investment becomes a wager placed without a scoreboard.
AI has lowered the administrative cost of maintaining that scoreboard considerably. Downtime classifications that once required an operator to pause, assess and manually log the cause of each event can now be captured and categorized without human intervention. Cycle times, production volumes and quality indicators feed into OEE calculations continuously, independent of whether individual teams have the bandwidth or the consistency to record them by hand. The data becomes both more granular and more trustworthy, which sharpens every downstream decision it informs.
Manufacturing also occupies an unusual position relative to other industries when it comes to validating AI expenditure, because the returns present themselves in units that finance teams can directly audit, such as incremental output per shift, recovered availability per reporting period and year-over-year reductions in maintenance cost against a documented baseline. Building the business case before deployment turns modernization from a capital gamble into a controlled experiment with verifiable outcomes.
Even the world’s most sophisticated AI tools will underperform if the team deploying them does not understand the operational context. How do operators interact with equipment under pressure? How do maintenance teams prioritize competing demands? How does downtime cascade across interconnected production lines? These are answers no algorithm can replicate and that no amount of sensor data can substitute for.
Manufacturers that can maintain accountability across the full lifecycle, rather than handing off between consultants and implementers, tend to deliver programs that stay on timeline and within budget.
The grace period for dealing with aging control systems is over. AI has made modernization faster, less costly and more measurable than ever before when paired with manufacturing expertise and rooted in data-backed business cases.
Manufacturers that embrace AI adoption now will set the competitive terms for the next decade. The ones still debating whether their 20-year-old systems can hold out a little longer are placing an increasingly risky bet.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。