At the SAP Sapphire conference, a number was announced: SAP will deploy over 200 AI agents across finance, procurement, supply chain, HR, and other full lines. Joule becomes SAP's unified entry point. AI is no longer just an assistant but a "process participant"—decisions can occur within the process, or even be directly executed by AI.
Around the same time, Odoo made a different strategic choice: natively embedding Anthropic's Claude into its chat module, while using the MCP open protocol to allow AI to directly read and write online business data. Instead of building a closed ecosystem, it took an open connectivity approach.
One uses a closed ecosystem to forcefully promote AI agents, while the other uses an open protocol to allow AI free access. These two paths convey the same meaning—ERP in 2026 is evolving from a "process recording tool" toward an "intelligent decision hub."
$89.7 billion IDC predicts global intelligent ERP market size in 2026, a year-on-year increase of 23.6%
Two Directions of ERP AI Transformation
SAP and Odoo represent two extremes respectively.
SAP's approach: closed-loop within the system. Joule acts as a "gateway," with all AI interactions conducted through it. Over 200 AI agents cover scenarios ranging from financial reconciliation to supply chain alerts. The benefits are deep integration, security, control, and seamless processes. The cost is—you must first move to the cloud (RISE with SAP) and migrate your system to SAP Cloud before you are eligible to use these AI capabilities. For long-time customers still on ECC, this essentially means, "If you want to use AI, first spend a year replacing your system."
Odoo's Route: Open Connection. Instead of building its own AI, it "plugs" mainstream AI tools like Claude, ChatGPT, and Gemini into Odoo through the MCP protocol. It does not restrict AI brands, does not require cloud migration, and data can be deployed locally. The advantage is flexibility and low startup costs—you can even start using the MCP Server with Odoo 14. The downside is that security configuration and permission management are your own responsibility.
There is no absolute right or wrong between the two directions; it depends on the company's own situation. However, for small and medium-sized enterprises and rapidly changing growth-stage companies, Odoo's open approach is clearly more attractive—it allows ERP to "enter the AI world" without first investing millions in cloud migration.
Three Structural Changes in the ERP Market by 2026
Setting aside the dynamics of individual vendors, the entire ERP industry has seen three major changes this year catalyzed by AI.
Change One: From "Process Online" to "Decision Online". Over the past two decades, the core value of ERP has been "moving business processes online"—procurement has documents, inventory has records, and finance has vouchers. With the addition of AI, the mission of ERP has changed: it is no longer just about recording what happened, but about predicting what is about to happen and then automatically deciding how to respond. IDC calls this the "ERP intelligent paradigm shift." The global intelligent ERP market is expected to reach $89.7 billion this year, with over 41% of the budget invested in AI function development.
Change 2: AI shifts from a "plug-in module" to a "native architecture". In 2026, over 50% of ERP products have adopted AI-native architecture—large models are directly embedded in the system's underlying layer, rather than just connecting an external API. Local vendors like Dingjie, Yonyou, and Kingdee are leading the way, achieving AI response latency below 0.3 seconds. SAP and Oracle follow a "platform + AI module" approach, with latency between 0.5 and 1 second. Those still using external AI plug-ins have clearly fallen behind.
Change 3: Productivity premium from local vendors' on-the-ground implementation. This is not a politically correct statement, but actual data from CCID Consulting for the first quarter. In manufacturing scenarios, vendors deeply rooted in the industry, such as鼎捷 and Infor, have seen AI implementation effectiveness improve by 35%-52%; international general-purpose vendors only achieved 15%-25%. In financial scenarios, the AI implementation effectiveness of Yonyou and Kingdee improved by 65%-80%. The gap is structural in fields like manufacturing, which demand extremely high "scenario understanding."
Differentiation Advantages of Open Source ERP
In this wave of ERP AI transformation, open-source ERP has gained an unexpected competitive advantage.
As SAP particularly emphasizes, enterprise AI requires "process knowledge, industry context, trustworthy data, and governance capabilities" to operate safely in critical business operations. SAP has 50 years of accumulated process knowledge, which is its moat. However, the advantage of open-source ERP lies in the fact that its data is completely open, allowing AI tools to freely "learn" from this data without needing to go through a closed API interface.
In plain terms: SAP's AI can only be used within SAP's system. Odoo's AI can connect to Claude, GPT, Gemini — you choose whichever works best for you. More importantly, the data stays on your own server, and it can be used even without the cloud.
Odoo's DACH market data for Q1 this year also supports this trend: revenue in the German market grew 133% year-over-year in the first five months, and the number of partners increased by 70%. Monthly new customer additions doubled, and DACH officially surpassed France to become the second-largest market in EMEA. The combination of open source and open AI integration is driving a new wave of ERP adoption.
ERP Selection: New Decision Dimensions
Previously, when selecting an ERP, the main considerations were whether the functional modules were comprehensive, whether the industry adaptation was good, and whether the price was suitable. In 2026, you need to add one more dimension: the depth and flexibility of AI integration.
I have compiled a simplified comparison table to help everyone understand the differences among the current mainstream options.
| Dimension | SAP | Odoo | DingJie/Yonyou |
|---|---|---|---|
| Architecture | Platform+AI Module | Open source optional + open protocol (MCP) | AI-native embedding |
| AI access method | Joule Unified Gateway | Claude native integration + MCP any AI | Self-developed industrial large model |
| Deployment Requirements | Must Go Cloud (RISE) | Available both locally and in the cloud | Support localized deployment |
| Startup costs | High (millions) | Low (Free Community Edition + Optional Enterprise Edition) | Medium (starting from 150,000) |
| Best Fit | Large multinational conglomerate | Small and medium-sized growth enterprises | Manufacturing/Central State-owned Enterprises |
| AI expands the degree of freedom | Limited to the SAP ecosystem | Open, not limited to any AI brand | Limited to self-developed ecosystem |
Three Suggestions for Business Managers
Based on these trends, I have three practical suggestions for managers who are planning digitalization.
First, don't wait for a "perfect solution".The integration of AI and ERP is still rapidly evolving, and there is no fully mature solution available right now. SAP's 200 agents are still running, and Odoo's MCP Server has only just over 20 users. But if you don't start trying, you'll always be stuck at the first level.
Second, start with data quality.Whether closed-source or open-source, AI always relies on data. Duplicate customers, inconsistent codes, incomplete purchase records—if these are not cleaned up, any AI solution will be useless. The "cleaning" I refer to is not a one-time effort—a continuous data quality mechanism should be established.
Third, focus on "openness." If you are implementing ERP now, this choice may affect the next 5-10 years. Choosing an open platform that does not restrict AI is far more flexible in the long run than choosing a closed vendor that only allows its own AI. This is the structural reason behind Odoo's growth in this wave—not because its features are superior, but because it allows enterprises to retain the freedom of choice.
The integration of ERP and AI, 2026 is just the starting point. By 2027 and 2028, when AI agents can automatically collaborate across systems and organizations, whether you choose "open connectivity" or "closed ecosystem" today will be amplified into a structural gap.
