For organizations, agentic AI represents a far bigger shift than just another new language model. It’s about how work is organized and how business processes are built, but also what role the ERP system takes on in an AI-driven organization.
First wave: Digitization laid the foundation
The first wave wasn’t really about AI. Over several decades, organizations digitized their processes. Financial systems, ERP, CRM, and other business systems replaced manual routines and created structured information. The result was enormous amounts of business data. It was only when this data started to become accessible that modern AI became possible.
Second wave: Generative AI
When large language models broke through, expectations of AI changed overnight. Suddenly, AI could
- write reports
- analyze documents
- create code
- translate
- summarize meetings
- generate images
AI became a powerful tool for knowledge work, but it still worked on request. A human asked for something. AI delivered.
Third wave: Agentic AI
Now the next shift is happening. Instead of helping people do work, AI is starting to do the work itself. This is what is commonly described as agentic AI.
An AI agent can, for example
- analyze incoming orders
- check inventory levels
- create purchase recommendations
- contact suppliers
- update the ERP system
- notify responsible parties if something deviates
without each step needing to be initiated by a human.
In other words, AI is no longer just a tool. It becomes a digital coworker.
From “Human in the Loop” to “Human above the Loop”
McKinsey describes the development as people increasingly working “above the loop,” i.e., further along in the flow and taking more of a supervisory role. That doesn’t mean people disappear from decision-making. Rather, the human role shifts. Instead of performing every task, the work is about
- setting goals
- defining rules
- prioritizing
- approving exceptions
- following up on results
AI does the operational work. Humans lead the process. That’s a much bigger change than simply using ChatGPT as a writing aid.
That’s why ERP becomes more important than ever
When AI starts executing business processes, a crucial question arises: How does AI know what’s right? The answer is rarely in the language model itself. It’s in the business’s processes.
The ERP system contains
- business rules
- master data
- authorizations
- processes
- financial relationships
- suppliers
- customers
- inventory
- purchasing
The more AI is allowed to do, the more important the quality of this information becomes. That’s why the ERP system is taking on a new role—not just as a transaction engine, but as AI’s operational brain.
Next step: Physical AI
Many observers believe the next step has already begun. When agentic AI is combined with robotics, sensors, and autonomous machines, what’s often called Physical AI emerges.
AI then leaves the computer and starts controlling the physical world. Examples already exist in
- autonomous warehouses
- self-driving vehicles
- robotic manufacturing
- drones
- logistics
What does this mean for organizations?
The biggest mistake is to think AI is mainly about which language model you choose. The real competitive advantage will likely be determined by something entirely different.
Organizations that have
- clear processes
- high data quality
- shared definitions
- a modern ERP platform
- clear data ownership
- effective governance
will be able to implement agentic AI much faster than organizations that still work in silos.
The future of AI is therefore less about technology than about the business. And perhaps that’s exactly why the ERP system—which many thought would become less important in the AI era—is instead becoming more important than ever.
FAQ
What is AI’s third wave?
AI’s third wave refers to the shift from generative AI to agentic AI. While generative AI primarily creates content—such as text, images, and code—agentic AI can plan, make decisions within given boundaries, and execute entire workflows. The focus shifts from AI as a tool to AI as a digital coworker that collaborates with people and other AI agents.
What is agentic AI?
Agentic AI consists of AI agents that can receive goals rather than individual instructions. They can analyze information, choose the next step, use multiple systems, and carry out tasks with limited human involvement. Examples include AI that handles procurement, customer service, or inventory planning by working directly in the organization’s ERP system.
What’s the difference between generative AI and agentic AI?
Generative AI creates content based on an instruction from the user, such as a text or an image. Agentic AI, on the other hand, can plan and carry out multiple steps on its own to reach a goal. Where generative AI answers a question, agentic AI can execute an entire workflow—for example, registering an order, checking inventory levels, ordering goods, and informing relevant people.
Why does ERP become more important as AI evolves?
AI needs to understand how the business works in order to make the right decisions. The ERP system contains business rules, processes, master data, authorizations, and transactions that give AI the business context it needs. The more AI automates the business, the more important data quality, process maturity, and a modern ERP system become.
How is SAP working with agentic AI?
SAP’s strategy is built around making the ERP system the hub for AI. With SAP Business Data Cloud, Joule, and AI agents, business data, processes, and AI are connected. The goal is for AI to work directly in business processes and use the same rules, data, and business context as the organization’s employees.
How is Microsoft Dynamics working with agentic AI?
Microsoft is evolving Copilot from a personal assistant into a platform for AI agents. Together with Dynamics 365, Microsoft Fabric, Dataverse, and Copilot Studio, organizations can create AI agents that work in business processes and use information from both Microsoft’s own systems and external data sources.
What is Physical AI?
Physical AI means AI controls physical systems such as robots, autonomous vehicles, warehouse robots, or drones. Here, AI is combined with sensors and robotics to perform work in the real world. Many see this as the next step after agentic AI, where digital decisions are turned into physical actions.
Do companies need to replace their ERP system to use agentic AI?
Not necessarily. But organizations with modern, cloud-based ERP systems often have better prerequisites because they offer standardized processes, higher data quality, open APIs, and built-in support for AI. For older systems, the challenge is often more about data, integrations, and process maturity than the AI model itself.
