Architecting the Foundation for Phase 2: From Rigid Belts to a Clean Digital Core

Part 1 talked about how pulp and paper companies are investing in AI but struggle to realize transformative returns, often due to legacy systems and rigid workflows that limit the technology's impact. We drew a parallel to the industry's historical shift from steam engines to decentralized electric motors, emphasizing that true productivity gains come from redesigning operational architecture—not just swapping out old technology for new. Now lets continue the discussion!

If we agree that the challenge is architectural, then the solution must be as well. The process rigidity we described is a direct symptom of an underlying IT architecture that has become inflexible over time. To move beyond the limitations of Phase 1, we have the opportunity to thoughtfully upgrade this operational drivetrain. This requires a strategic approach that resolves a seeming paradox: how to gain more flexibility by embracing standardization.

The journey begins by modernizing the heart of the system: the ERP. The solution lies in transitioning from a complex, heavily customized on-premise ERP to a cloud-based ERP guided by the principle of a "clean core," a pathway enabled by solutions like SAP Cloud ERP. A clean core strategy means keeping the central ERP system as close to standard as possible. This standardization is what makes the core robust, secure, and, most importantly, simple to upgrade. It ensures the system can always incorporate the latest innovations—including new AI capabilities—without the risk and expense of massive, multi-year projects. It prevents the core from becoming a monolithic bottleneck.

So, where does the flexibility come from? It comes from decoupling customization and innovation from that stable, standardized core. This is the role of the SAP Business Technology Platform (BTP). Instead of altering the core to meet a unique business need, you build a new application or extension on BTP. This new piece of functionality then integrates seamlessly with the core system but runs separately. This "composable" architecture is the modern equivalent of replacing the rigid drive shaft with flexible electrical wiring. It provides immense agility, allowing teams to develop and deploy new capabilities and AI services rapidly, without destabilizing critical business operations.

The third, and perhaps most critical, piece of the puzzle is ensuring we have high-quality fuel. AI runs on data, and a primary reason for disappointing project outcomes is a lack of clean, reliable, and accessible data. The SAP Business Data Cloud, featuring SAP Datasphere, is designed to address this by creating a "business data fabric". This technology provides a unified way to access, integrate, and model data from both SAP and non-SAP sources, wherever it may be. Crucially, it preserves the business context and logic of that data, ensuring that the information feeding our AI models is not just raw data, but meaningful, reliable, and ready for analysis. 

Together, these three elements form a holistic strategy that addresses the primary challenges of AI adoption: process rigidity, a lack of agility, and poor data quality.

Unleashing Phase 2: The Intelligent Mill Powered by SAP Business AI

With this modern, flexible, and data-rich foundation in place, the conversation can shift from optimizing isolated tasks to orchestrating intelligent, end-to-end processes. This is the promise of Phase 2. It's where SAP Business AI—the intelligence embedded across the entire SAP suite—can begin to operate at its full potential.

The user experience in this new phase is transformed by tools like Joule, SAP's AI copilot. Joule is designed to understand the business context of a user's request and can help execute complex tasks across multiple applications using natural language. This is powered by a network of specialized AI agents that can autonomously manage multi-step processes. These agents are designed to collaborate with each other to help resolve complex, cross-functional issues, such as a supply chain disruption or a customer payment dispute, with greater speed and efficiency.

For us in the pulp and paper industry, this represents the difference between predicting a single equipment failure and orchestrating an intelligent response to a market shift.

In a Phase 2 scenario, an autonomous supply chain agent could anticipate a raw material shortage, simulate its impact on production, and recommend adjustments to production plans in real-time to minimize disruption, all while communicating the business impact to the relevant teams via Joule. This is the future that AI promises, and it's a future that becomes possible on a modern architectural foundation.

Conclusion: The CIO as Architect, with a Human at the Helm

The journey to unlock the full return on our AI investments seems to be a journey of architectural evolution, not just technological adoption. This positions the CIO not just as a manager of technology, but as a strategic architect, designing the intelligent enterprise foundation of the future. It's a business transformation that calls for a clear vision and collaborative leadership.

As we architect this future, it's reassuring to know that the goal is not to replace human expertise, but to augment it. SAP's philosophy is built on the principle of keeping the "human in the loop".

AI is at its best when it serves as a powerful collaborator for our teams. It is designed to automate the complex and repetitive, freeing our experts to focus on strategic oversight, creative problem-solving, and the next wave of innovation. This approach, where final authority rests with people, not only mitigates risks but also builds the trust and acceptance within our workforce that is so essential for any successful transformation.

The time feels right to move beyond the cycle of incremental gains and pilot projects. The technologies and the strategic framework are now available to build the foundation for the next generation of our industry. It seems the opportunity is here to stop tinkering with the transmission belts of the past and start architecting the intelligent, autonomous mill of the future.

To explore how SAP can partner with you in designing this new foundation with a clean core, an agile innovation platform, and a unified data fabric, go here: and for more about how SAP supports the mill products industries including paper and packaging go here.

I hope you saw the October 12 episode of “60 Minutes”. If not, you should find it and watch it.

The episode includes a story about a water treatment plant in the small town of Littleton MA that got hacked by China. They are not alone. According to the FBI, 200 similar facilities had been hacked. While hacking your company network is very damaging and costly, hacking an industrial control system can be deadly. When an unfriendly adversary can have control of valves and motors, they can shut down operations or cause deadly accidents. The story included an interview of 4 star General Tim Haugh who was leading the federal effort to defend the United States from computer threats. General Haugh said “If you are willing to go after a small water provider in Littleton, Massachusetts, what other target is off the list?”

You are on the list. Every company has an Information Technology (IT) department that tries to protect their network, but industrial control systems are different. You cannot use the same cyber security approach for IT as Operational Technology (OT). The objectives and technologies are very different. If you don’t have a plan and personnel to specifically address the cyber security threat to your industrial control system, you may find yourself weeping and wondering why you didn’t act before it happens to you.

That is why the International Society of Automation (ISA) has a certification for industrial cyber security. The 62443 standard is designed for the unique nature of threats to Supervisory Control and Data Acquisition (SCADA) and Distributed Control Systems (DCS). 

Applying this standard requires people that know your system. You need to know Rockwell, Emerson, Honeywell, ABB, Siemens, Valmet, Inductive Automation, Schneider, and other platforms. Each one has its way of doing things and requires a competent person on that platform to implement hardening. An example of cyber vulnerability is that often user interface scripting performing in Python or Visual Basic can have vulnerabilities if not implemented carefully. It not only requires competency in the programming language but also has to consider how your industrial control platform works. 

I have seen firms that assumed their IT department could take care of it. In one case, the corporate IT implemented a company wide firewall approach. Because IT did not know anything about OPC communications, this resulted in communication losses on the control network. Very dangerous!

If I can get you to take one action after reading this, find “China is hacking America's critical infrastructure, former NSA and retired Gen. Tim Haugh warns” and watch the “60 Minutes” story. It should scare you. Your next action should be to find an ISA 62443 certified resource to protect your industrial control system.

As Kai moves into the vision of what I will call "Advanced AI" in the second part of his 2-part article above, one of the challenges I see is presenting such matters in such a way that the human "AI Architects" can mentally visualize them.

I first experienced this with data bases when I work for Jaakko Poyry. We were building an engineering design basis. I could get 2 or 3 dimensional data bases. I could even get 4 (add the dimension of time). But 5, 6, 7 or more dimensional data bases? Couldn't visualize them. AI stretches us to visualize issues that do not exist in the natural world.

Since all matters end up being used in the real world, it is going to be the responsibility of the "AI Architects" to translate what they are doing for us in the real world. A formidable challenge.

After years of pilots and proofs of concept, private 5G networks are now operating inside real manufacturing environments across Europe, the U.S. and Asia. Automotive plants, semiconductor facilities, machinery workshops and advanced logistics hubs are among the first to run production systems on private 5G rather than wired or Wi-Fi-only infrastructure. These deployments demonstrate both the genuine advantages of the technology and the practical challenges factories must confront when scaling from trials to live operations.

For decades, the production of wood materials, even high-specification products like plywood and laminated veneer lumber (LVL), lagged far behind the technological adoption seen in the automotive, steel, and semiconductor industries. Production relied heavily on manual labor, visual inspection, and batch-processing—a methodology inherently prone to material inconsistency.

From AI-powered detection to cloud strategies, industrial companies are using multi-layered defenses against ever-evolving digital threats and operational technology exploits.

Manufacturing executives are wagering nearly half their modernisation budgets on AI, betting these systems will boost profit within two years.