Manufacturing’s AI Race Has a Workflow Problem
Across semiconductor fabs, pharmaceutical plants, and data centres throughout Asia, a familiar pattern is playing out. Leadership teams, under pressure to show they are “doing something” with AI, greenlight pilot after pilot. Chatbots evolve into agentic systems. Even with allocated budgets and engaging vendors, for many organisations, the return of investment never quite materialises
The problem, according to Kenneth Sim, Chief Technology Officer of OneSystems Technologies, isn’t the technology itself. It’s the sequence in which companies are adopting it.
“Many companies are jumping into it and following all the various buzzwords being thrown around everywhere,” Sim explains. The shift from traditional chatbots into agentic AI has been rapid, but he’s observed that “we hardly notice customers that focus on workflow” before deploying it. The result: organisations “burst through their AI budget very quickly,” chasing autonomy before they’ve defined the process the AI is meant to optimise.
For an industry where margins are unforgiving and downtime is measured in millions, this is more than an inefficiency. It’s a strategic risk, and the kind of challenge that leaders across the region’s advanced manufacturing sector are now being forced to confront.
Meet the Man Bridging Two Worlds
Kenneth Sim has spent two decades building OneSystems Technologies into a company that operates in a space few competitors are equipped to occupy: the intersection of Operational Technology (OT) and Information Technology (IT). His role, he says, sits “typically at the intersection between technical strategy as well as delivery,” overseeing architecture, engineering, and the integration of emerging technologies for forward-looking customers.
OneSystems focuses on three sectors, semiconductor, pharmaceutical, and data centre industries, with a specialisation in security and communications infrastructure. What distinguishes the company, Sim notes, is that it has never outsourced its technical capability. Every competency, from architecture to delivery, has been built in-house over 20 years.
In an era where AI vendors and system integrators are proliferating rapidly, Sim points out that many competitors “subcontract in certain phases of their work,” which introduces risk that clients often don’t see until something breaks. Vertical control, he argues, is what allows OneSystems to “produce consistency,” “continue to innovate with the customer,” and “stay accountable with the customer for the long term.”
The Hidden Problem: AI Without Workflow
The clearest thread running through Sim’s perspective is this: AI amplifies whatever process it’s layered onto, good or bad. Deploy it against an undefined or undocumented workflow, and you don’t get intelligence. You get expensive noise.
He’s watched organisations take the path of least resistance, testing AI in “the least crucial departments,” such as HR, to “see how it goes.” Although the initiative feels safe, it also has limited business impact. The organisations seeing genuine returns, by contrast, take a harder but more deliberate path, auditing where their data actually lives, then realigning their data strategy to support AI functions that drive measurable business value.
His recommendation for manufacturing leaders evaluating where to start is straightforward:
Identify your organisation’s largest operational pain point
Document your workflow before automating it
Build AI agents around the workflow
Expect faster ROI once the sequencing is correct
This is a meaningfully different mental model from “let’s try AI and see.” It reframes AI adoption as a discipline of industrial engineering rather than a software rollout, a distinction that separates manufacturers accelerating their operations from those quietly absorbing sunk costs.
Why OT-IT Convergence Is No Longer Optional
If workflow discipline is the first pillar of Sim’s philosophy, OT-IT convergence is the second, and arguably the more structurally important of the two for advanced manufacturing specifically.
Historically, Operational Technology (the systems automating and monitoring physical equipment) and Information Technology (the systems governing data and communications) have existed in silos, built by different teams, on different timelines, with different priorities. OneSystems has built its entire go-to-market strategy around closing that gap.
“What we observe is that we actually wedge ourselves between OT as well as IT,” Sim says, describing OT as fundamentally about the automation and monitoring of physical equipment, while IT governs data communications and bandwidth, a distinction that has become sharper as AI workloads spread across factory floors. The differentiator, he says, is that OneSystems has spent two decades learning to “speak languages of both IT as well as OT”, a bilingual fluency that most integrators, and most in-house IT teams, simply don’t have.
For manufacturing executives, particularly CIOs and Engineering Heads, this convergence isn’t a future-state ambition. It’s already the operating reality of any facility running AI-enabled equipment, next-generation Manufacturing Execution Systems (MES), or connected sensor networks. Sim notes that OneSystems partners with MES technology providers to help tune the AI already embedded in their products, refining how those systems process data on the ground, in real manufacturing environments, so the tools genuinely improve rather than simply generate dashboards.
Clean Data Is the Real AI Infrastructure
According to Sim, companies should first invest in data hygiene prior to investing in AI. “The faster, the cleaner the data, the more accurate AI can be,” he says. “It’s always garbage in, garbage out.” Before OneSystems can even begin deploying meaningful AI capability for a client, Sim notes that “half the time before we get there, we need to first tidy up their data sets.”
His advice to manufacturing leaders is almost disarmingly practical: audit your spreadsheets. Behind that simplicity is a serious point, most manufacturers don’t have an AI readiness problem, they have a data governance problem, and no amount of sophisticated modelling will compensate for inconsistent, siloed, or poorly structured operational data.
For functions like Quality Control, R&D, and Production, this has direct implications. AI-driven defect analysis, predictive maintenance, and yield optimisation are only as good as the sensors and process data feeding them.
Cybersecurity: The Convergence Blind Spot
As factories become more connected, the attack surface expands in ways many security teams aren’t yet accounting for, and this is where Sim’s OT-IT thesis becomes a cybersecurity argument as much as an efficiency one.
He describes a structural blind spot: cybersecurity specialists typically focus on network traffic and IT-layer vulnerabilities, but rarely ask about “the default password for your video management system.” That system, however, is connected to the same IT backbone, meaning a weak point in a physical security device can become a backdoor into core infrastructure.
As AI moves to the edge, embedded directly in devices performing specific functions on the factory floor, the number of these vulnerable entry points multiplies. Sim also raises a subtler risk that deserves more attention from Compliance, Audit, and CISO functions: verifying that edge-collected data is genuinely clean, with no false data injection compromising the integrity of what AI systems are learning from.
For high-value manufacturing environments handling sensitive IP, Sim’s conclusion is unambiguous: OT and IT security cannot be managed as separate disciplines. They need to be “married together in one seamless, manageable platform that is scalable.”
AI Video Analytics: Beyond Surveillance
One of the more immediately actionable applications Sim points to is AI-powered video analytics, a category he’s careful to distinguish from traditional CCTV.
Rather than static surveillance, computer vision paired with AI now supports real-time operational safety. Sim gives a simple, relatable example: workers walking through a facility while looking at their phones, at risk of colliding with equipment or colleagues. AI-enabled monitoring can flag unsafe behaviour as it happens, prompting a response immediately, rather than relying on staff to recall standard operating procedures from memory.
The bigger shift, he argues, is from a security function to a productivity function. Applied to yield analysis, this same principle allows manufacturers to identify the source of a production line defect, even one as small as 1%, by analysing centralised sensor data across the entire line, compressing what once took “half a day” of manual investigation into a matter of minutes.
Workforce Readiness: Prompt Engineering Isn’t Optional
Sim is direct about where the skills gap actually sits, and it isn’t where most executives assume.
Rather than needing to build deep AI engineering capability in-house, he argues manufacturers need two more foundational competencies: workflow documentation and prompt engineering. On the latter, he’s blunt: “It’s very easy to use any open source AI, you ask the AI ten times, you probably get five different answers, because you didn’t give it context.” Prompt engineering, in his view, is a genuine skill, one requiring specificity, context, and a healthy scepticism toward AI outputs, particularly from open-source tools.
His broader philosophy on AI autonomy reflects a disciplined, risk-aware approach well suited to regulated, high-precision manufacturing environments: implement AI with continuous human oversight, review the architecture consistently, and resist treating deployment as “fire and forget.” Using a familiar approvals framework, he suggests aiming for AI to responsibly handle “level 1.5” decisions, automating elementary, repetitive judgment calls while keeping complex, high-stakes decisions firmly with human leadership.
What Comes After AI: Quantum and Dark Factories
Looking beyond current AI deployment cycles, Sim points to quantum computing as the next major inflection point, one that could compound the impact of AI already in production. Greater compute power, he suggests, will accelerate AI’s capabilities further and enable “truly autonomous, full autonomy of operations,” including the emergence of fully autonomous “dark factories.”
His timeline estimate for the region is notably near-term: if quantum computing accelerates as expected, he anticipates seeing more dark factories emerge across ASEAN within three to five years.
On robotics, Sim is more measured. While investment in autonomous and humanoid robotics is accelerating, he sees the technology as still maturing, humanoids today are “programmed to perform a few functions” and don’t yet operate with human-like adaptability. That said, he expects the technology to evolve steadily, gradually taking on more functions currently performed by people.
Why Ecosystem Collaboration Matters More Than Ever
Perhaps the most instructive part of the conversation is Sim’s framing of OneSystems’ own success. Despite building deep in-house capability, he’s emphatic that no single company can do everything alone. “Without our technology partners, OneSystems won’t be where we are today,” he says, describing his company’s role as an integrator that couples specialised technologies, such as AI video analytics from surveillance vendors, into complete operational solutions.
This is the same logic driving the broader industry conversation happening across semiconductor, electronics, medical tech, and chemical manufacturing sectors in Asia: complex challenges like OT-IT convergence, industrial cybersecurity, and workforce upskilling aren’t solved in isolation. They require manufacturers, technology providers, systems integrators, investors, and policymakers working from a shared understanding of where the industry is heading.
Join the Conversation at AIM Asia Week
Conversations like this one, grounded, specific, and free of hype, are exactly what the region’s manufacturing leaders need more of as they navigate AI adoption, OT-IT convergence, and industrial cybersecurity under real operational pressure.
That’s the purpose behind AIM Asia Week. The event brings together manufacturing leaders, government representatives, investors, innovators, and ecosystem partners from across semiconductor, electronics, medical technology, and chemical industries to shape the future of advanced manufacturing in Asia.
Kenneth Sim will be among the industry voices at AIM Asia Week 2026, continuing this conversation on OT-IT convergence, industrial AI, and cybersecurity with manufacturing executives across the region.
Register now to secure your place
Connect with AIM Asia
Join us at our upcoming AIM Asia programmes to connect with industry leaders, innovators, governments, investors, and technology providers shaping the future of manufacturing and industrial transformation across ASEAN. Connect with us to gain access to the exclusive AIM Circle platform and stay connected with industry leaders and opportunities across the advanced manufacturing ecosystem.
For AIM Circle access, speaking, sponsorship, exhibition, and partnership opportunities, contact [email protected] or +65 8868 1418 (SG) | +6011 1616 3281 (MY)