The pulp and paper industry has always been a game of razor-thin margins and massive physical scale. For decades, the Chief Information Officer’s primary mission was to maintain a stable digital mirror of these physical realities—ensuring that the flow of data from the forest to the finishing line remained uninterrupted. We optimized our ERPs, refined our MES layers, and more recently, experimented with predictive maintenance to keep the mill running. Now that we have entered 2026, the technological landscape is shifting beneath our feet. We are moving beyond the era of systems that merely record or suggest actions. We are entering the age of the autonomous enterprise, driven by Agentic AI.

In an industry defined by capital-intensive assets and extreme price volatility in energy and raw materials, this represents more than just another incremental software update—it is a fundamental evolution. Over the past few years, we have seen AI transition from Predictive AI—forecasting bearing failures—to Generative AI, capable of summarizing shift reports. While impressive, these capabilities often missed the mark for those on the front lines. For a mill manager facing a 2:00 AM web break or a VP of Operations monitoring real-time energy surges, a chatbot that writes poetry is hardly a game-changer.

Now, the pivot point is Agentic AI. This refers to systems that don’t just talk; they act. They coordinate. They solve problems across departmental silos without waiting for a human to trigger every individual step.

The Universal Adapter: Beyond the Limitations of APIs

To understand how this shift will manifest in a paper mill, we must first look at the underlying plumbing. For years, the industry relied on APIs—Application Programming Interfaces—to connect the shop floor to the top floor. However, APIs have a built-in limitation: the caller is responsible for everything. If you wanted to connect a logistics system to a production schedule, the developer had to manually map every parameter, ensuring the "handshake" was perfect. It was rigid and fragile.

This is where the Model Context Protocol, or MCP, comes into play. Originally introduced by Anthropic and now being championed by leaders like SAP, MCP is essentially the "universal adapter" for the age of AI. Unlike a traditional API, where the caller must know the exact structure of the data, MCP allows an AI agent to understand a tool’s capabilities through an "agent card." It describes what it can do and what it needs, allowing the agent to figure out the "how."

For the CIO, this means an architecture upgrade is imminent. By the first quarter of 2026, the SAP Integration Suite will feature an MCP Gateway. This solution acts as the foundation for connecting cloud and on-premise applications with pre-built content, effectively becoming the bridge that allows your legacy mill systems to speak the language of modern AI agents. When paired with the SAP Business Data Cloud—a comprehensive data service that ensures mission-critical business data is accessible and scalable—your agents finally have the high-fidelity context they need to make autonomous decisions.

A Network of Intelligence: A2A and the Human Element

One of the most common misconceptions about the future of AI is that it will be a single, monolithic "brain" controlling the mill. The reality is far more sophisticated. SAP is moving toward an architecture of specialized agents. Each agent is an expert in its narrow field—one for procurement, one for energy optimization, one for maintenance.

As part of our ongoing partnership with Google, SAP is leveraging the Agent2Agent (A2A) interoperability protocol, which establishes a foundation for AI agents to securely interact and collaborate across platforms. This protocol allows different agents to collaborate, exchange updates, and, crucially, ask questions. It is important to note that in this ecosystem, an "agent" isn't always an AI. An agent can be a legacy software program, a robot on the warehouse floor, or a human supervisor. The A2A protocol ensures that if an AI agent reaches a point of uncertainty, it doesn't just halt the process or make a "hallucinated" guess; it can reach out to a human colleague for clarification or approval, keeping the "human-in-the-loop" as a vital part of the workflow.

Joule: The Conductor of the Digital Symphony

At the center of this collaborative network sits SAP Joule. While many initially viewed Joule as a simple chatbot, its role has evolved into that of an orchestrator. SAP Joule is the AI-powered copilot that understands your unique business context, acting as the interface through which a CIO or a Production Manager directs the entire agentic workforce. Joule doesn't just fetch information; it triggers workflows across the specialized agents in SAP. As the digital heart of the enterprise, SAP provides the real-time process data and financial transparency that these agents require to ensure their autonomous actions align with the company’s bottom line.

From Theory to the Shop Floor: The Smart Drying Scenario

To see the tangible value of this architecture, imagine a future "Smart Drying Scenario," a challenge every paper manufacturer faces. Drying is the most energy-intensive part of the process, and in an era of volatile energy markets, timing is everything.

In the near future, the process could begin not with a human checking a screen, but with an external Market Intelligence Agent. This agent monitors energy spot prices in real-time. It detects a projected price spike for the following afternoon and immediately informs SAP Joule. Joule doesn't wait for a morning meeting; it triggers a simulation within the SAP Production Planning Agent.

The system simulates a revised schedule: What if we accelerate the production of energy-intensive, heavy-weight linerboard (requiring deep drying to 5% moisture) tonight, and switch to less intensive medium-grade orders during tomorrow’s price peak? This would cater to a local converter that accepts a 'shipping moisture' of 7.5%.

Simultaneously, a Logistics Agent checks the warehouse capacity and shipping schedules to ensure this shift won't cause a bottleneck or delay a high-priority customer delivery. Within minutes, Joule presents the optimal revised plan to the Production Manager. The manager sees the projected savings—thousands of euros in avoided energy costs—and the confirmed feasibility from logistics. With a single click of approval, the agents automatically update the production schedules, notify the warehouse team, and adjust the raw material procurement orders. The mill continues to hum, the customers remain satisfied, and the margin is protected—all without a single manual spreadsheet calculation.

A New Mandate for the Workforce

This shift in technology inevitably brings a shift in the role of the people who run our mills. We are moving away from a world where employees spend their days executing repetitive processes—entering data, reconciling reports, or manually adjusting schedules. In the agentic era, the human role becomes one of "Intent Orchestration."

Your engineers and planners will no longer be defined by their ability to operate the software, but by their ability to define the goals. They will set the parameters—sustainability targets, quality benchmarks, and cost limits—and the agentic system will figure out the most efficient path to reach them. This elevates the work of the mill staff, allowing them to focus on strategy, innovation, and handling the complex exceptions that truly require human intuition.

Conclusion: The Window is Opening

The era of Agentic AI is no longer a topic for a five-year roadmap. With the upcoming 2025 and 2026 releases of SAP’s core technologies, it is becoming an operational reality. For the pulp and paper industry, the opportunity is clear: to break through the efficiency ceilings of traditional automation and build a mill that is as resilient as it is productive.

The question for the CIO is no longer whether to adopt AI, but how quickly you can prepare your data and integration layers to support this new workforce of agents. The tools are ready, the protocols are set, and the first movers in the industry are already beginning to orchestrate their future.

Are you ready to lead your organization into the next phase of autonomy? Explore how SAP is shaping the future of industrial intelligence and start building your own network of specialized agents today.

Visit the SAP Agentic AI Hub

The International Society of Automation (ISA, www.ISA.org) is the world authority on automation. Therefore, an ISA position paper entitled “Industrial AI and Its Impact on Automation” should carry some weight.

A few notable points made in this paper are:

• Artificial Intelligence (AI) has been used in automation since the 1960s. My guess is most people reading this don’t believe this, and certainly the general public thought AI began with the introduction of publicly available free ChatGPT 3.5 in 2022.

• The paper tells us to be very careful about applying AI in industry.

• If badly implemented, AI can make industry less safe. To quote “Widespread and uncontrolled adoption of AI to automate activities that were previously the province of human experts introduces new vulnerabilities, creating new attack surfaces where automation is used for the first time.”

• Industrial automation requires high availability and reliability, and any AI application cannot be allowed to degrade those requirements.

• AI requires large volumes of high-quality data, but industrial processes have noise and outliers.

• Generally, AI techniques are black box. We don’t necessarily know what is going on inside its digital brain. In a rather provocative statement, the paper states “The technology is immature and poorly understood, with no well-established assurance regime and a lack of clear standards.”

• AI can introduce cyber security risks.

• AI implementations must comply with industry standards, regulations and legal requirements.

• AI introduces new ways of operating which require training of personnel to use it properly and safely in industry.

The bottom line is that the world authority on industrial automation wants to express caution amid an AI boom. There is heavy investment in AI, and it could be a bubble that will burst. ISA does not want industry to dive into the deep end of AI without making sure there is water in the pool. The paper recognizes the huge benefits AI can make, and frankly those facilities that fail to keep technological pace will not be sustainable as their competitors surpass them. However, you need experienced and skilled people to properly utilize AI.

Unlike ChatGPT, our processes move valves and motors with chemicals and steam in processes like recovery boilers that can quickly cause huge financial loss and injury. Get people with the industrial automation experience required before you apply AI in your operation.

The developments are coming fast and furious. While Kai's description of Agentic AI is exciting, it leaves one with the question, "What will the humans do?" when this is fully developed.

I think humans are pushed further out in the future and spend more time in strategic functions than in tactical activities. Agentic AI will usurp all the tactical functions. Humans will be working on future strategies.

This may require a different expertise and skill set on the human level than we have now.

Decades ago, I had input to an article in National Geographic. I learned at that time that NG assessed every article idea they received in the following way. From the beginning through publication they were asking the question, will this article be relevant in two years? Right up to publishing they ask this question and if the answer was then turned to "no" the article was scrapped. Compare that to a newspaper editor who lives on a 24-hour cycle.

Our managers have been living on the 24-hour cycle. Agentic AI is going to move them to the two-year cycle. It will be quite a change.

The narrative in modern manufacturing often centers on the cutting edge: AI-driven robotics, hyper-connected IIoT ecosystems, and autonomous logistics.

Manufacturers enter 2026 under pressure to do more with less, but the path from firefighting to flow is becoming clearer as ERP and adjacent technologies mature. For technology leaders, the real change is not another dashboard, it is a shift toward predictable, stable operations built on connected, data-driven processes. 

Devices in a broad range of edge AI applications are increasingly at risk of hacking or tampering, with the stakes varying greatly depending on how much the device can impact and interact with human life. Design methods and protection techniques must now be included up front in the design cycle for optimal protection of consumers and companies as the quantum threat looms.

GAITHERSBURG, MD—The National Institute of Standards and Technology (NIST) has expanded its collaboration with the nonprofit MITRE Corp. as part of its efforts to ensure U.S. leadership in artificial intelligence technology. NIST is investing $20 million to establish two centers to advance AI-based technology.