XMPro

XMPro CEO Pieter van Schalkwyk speaking at CIIT Latam Congress 2026. His keynote focused on a shift industrial organisations can’t avoid: Moving beyond monitoring… toward decision-centric operations and governed autonomy. Because the real challenge isn’t building AI. It’s deploying systems that can be trusted in live industrial environments, where decisions carry operational, financial, and safety consequences. That’s where Agentic Operations becomes relevant. Not as a concept, but as a practical approach to: • Identify what matters in real time • Support and coordinate decisions • Operate within defined boundaries and governance Appreciate the opportunity to contribute to the conversation with CIIT Perú and the broader industrial community. #AgenticOperations #IndustrialAI #DecisionIntelligence #CIITLatam26

𝐗𝐌𝐏𝐫𝐨 𝐡𝐚𝐬 𝐛𝐞𝐞𝐧 𝐫𝐞𝐜𝐨𝐠𝐧𝐢𝐳𝐞𝐝 𝐛𝐲 LNS Research 𝐚𝐬 𝐚 𝐤𝐞𝐲 𝐯𝐞𝐧𝐝𝐨𝐫 𝐢𝐧 𝐭𝐡𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐎𝐩𝐞𝐫𝐚𝐭𝐢𝐨𝐧𝐬 𝐜𝐚𝐭𝐞𝐠𝐨𝐫𝐲. --- What’s notable here is the framing. Most industrial AI today still stops at insight... dashboards, alerts, predictions. Agentic Operations points to the next step. Systems that don’t just inform decisions, but coordinate and execute them. It’s a shift we’ve been building toward for some time, so it’s good to see it starting to take clearer shape in the market. — Read about the full landscape here: https://lnkd.in/gbPibzQh #AgenticOperations #IndustrialAI #AgenticAI #XMPro #LNSResearch

“If you could put a brilliant process engineer next to your operators 24/7… what would you ask them to do?” That question started a conversation that eventually became #BrewVerse. In this short clip we had a bit of fun with it — using an AI digital twin of XMPro’s Warren Jackson to explain the idea that kicked the whole project off. But the project itself is very real. At NVIDIA GTC, the BrewVerse demo shows how live manufacturing data streaming into NVIDIA Omniverse, combined with multi-agent operational intelligence, can support operators by continuously monitoring process conditions, running calculations, and helping reduce variability in real operations. If you’re at GTC this week, come see it live at the @Dell Technologies booth. Huge credit to the teams from Dell Technologies, NVIDIA, New Belgium Brewing, LinkToVR, and New Frontier Technologies, and to Gavin Green, VP of Strategic Solutions at XMPro, for helping bring this together. And if you recognise the brewery asset featured in the digital twin, drop it in the comments 👀 #GTC2026 #NVIDIA #Omniverse #IndustrialAI #DigitalTwin #Manufacturing #XMPro #MAGS

𝐓𝐡𝐞 𝐓𝐞𝐧 𝐋𝐚𝐰𝐬 𝐨𝐟 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐢𝐚𝐥 𝐀𝐈 | 𝐋𝐚𝐰 𝟏 Over the coming weeks we’ll explore the Ten Laws of Industrial AI from The Industrial AI Agent Manifesto, a Digital Twin Consortium publication. Download the manifesto here --> https://lnkd.in/gEsCvx5z 𝐋𝐚𝐰 𝟏: 𝐃𝐞𝐭𝐞𝐫𝐦𝐢𝐧𝐢𝐬𝐭𝐢𝐜 𝐕𝐚𝐥𝐢𝐝𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐄𝐱𝐞𝐜𝐮𝐭𝐢𝐨𝐧 Industrial agents must produce deterministic validated actions. In other words, the same operational conditions must produce the same validated action. That sounds obvious, but it exposes one of the biggest gaps between consumer AI and industrial AI. Most general-purpose AI systems are stochastic. The same prompt can produce different answers each time. That behaviour is acceptable when the task is writing an email or summarizing a document. It is unacceptable when the system is involved in operational decisions affecting safety, production, or equipment. In industrial environments, identical operational states must lead to predictable and reproducible validated actions. This requires deterministic validation mechanisms that separate AI reasoning from execution. This does not mean AI cannot reason probabilistically. It means that before any recommendation becomes an operational action, it must pass through a deterministic validation layer governed by mathematical constraints, safety rules, and operational limits. This architecture allows AI to assist with complex decisions while ensuring predictable behaviour in safety-critical environments. Download the full manifesto here: https://lnkd.in/gEsCvx5z

Many AI and automation projects still start by asking the same question: what is the process? But when working with cognitive agent systems, that may be the wrong starting point. In this new article, Pieter van Schalkwyk explores why designing the decision environment matters more than prescribing the workflow, and how this shift changes the way industrial AI systems should be built. A useful perspective for anyone working with agent systems, digital twins, or industrial operations. Read the full article below.

A common mistake in AI agent design is starting with the workflow. In this article, XMPro CEO, Pieter van Schalkwyk, explains why that instinct often leads to systems that behave like expensive workflow engines rather than true cognitive agent teams. A thoughtful read for anyone designing industrial AI systems.

In this new article, XMPro Technical Consultant, Marcos Augusto Burgos Saavedra explores how causal analytics moves beyond predicting failures to identifying the upstream process conditions that create them. Predictive maintenance can warn when equipment is likely to fail. But many operations teams still find themselves replacing the same components repeatedly because the underlying conditions causing those failures remain unchanged. Causal analytics shifts the focus from reacting to symptoms to understanding the cause-and-effect relationships across the entire process chain. In mining operations, for example, a bearing failure in a flotation pump may be influenced by upstream variables such as grinding circuit performance, cyclone parameters, or slurry conditions. By uncovering these relationships, organizations can move beyond simply predicting failures toward eliminating the conditions that cause them. Read the full article: https://lnkd.in/gA8_rBBk #IndustrialAI #PredictiveMaintenance #CausalAI #DigitalTwins #Mining #XMPro

The Digital Twin Consortium has published the Industrial AI Agent Manifesto: a governance framework for deploying AI agents in safety-critical industrial environments. As AI agents move from experimentation into operational decision-making, a critical question emerges: Can you demonstrate what your agent decided, why it decided it, and how unsafe outcomes are structurally prevented? The manifesto defines ten engineering laws for trustworthy autonomous operations, including: • Deterministic validation and execution • Physics- and process-aware intelligence • Separation of cognition and actuation • Human override and emergency stop capabilities • Progressive autonomy with defined safety boundaries • Multi-agent safety orchestration • Safe and secure continuous learning This framework is intended for sectors where operational integrity matters: manufacturing, energy, mining, aerospace, healthcare, and building systems. The full manifesto PDF is attached to this post. If your organization is evaluating agentic AI for industrial operations, governance architecture is not optional. Autonomy must be engineered. #DigitalTwinConsortium #IndustrialAI #AgenticAI #DigitalTwins #AIgovernance #AutonomousOperations #XMPro

OpenClaw proved the market wants agents that act. It also proved, in under eight weeks, that unbounded execution is a direct path to compromise. Our team wrote about why industrial operations need a different architecture... and why we built one. 👉 Bounded autonomy. Bounded actuation. Governance from the ground up.