In the second week of July, I laid out three tasks on the table.
First: Odoo 20 is scheduled to be released at Odoo Experience in Brussels from September 24-26. In this major version update, AI is no longer a separate functional module but an "execution layer" embedded in every core application. Accounting AI performs automatic audits, inventory AI predicts replenishment, timesheet AI automatically records, and CRM AI continuously monitors the sales pipeline—AI has learned to "work on its own," no longer requiring people to wait for instructions.
Second: Jin Die Vice President Liu Zhongwen made a weighty statement in an interview: "The depth of this transformation far exceeds that of cloud transformation. Cloud transformation changes the delivery and deployment methods, while AI-native transformation changes the very operation of enterprises." Jin Die's answer is "Lingji," an AI operating system rebuilt from the ground up.
Third: This Thursday (July 16), Odoo will hold the "Digital Intelligence Innovation Day" in Shenzhen, specifically tailoring an AI-enhanced ERP solution for enterprises going global. With 70+ official applications fully integrated, 32 languages, 167 currencies, and financial compliance in 100+ countries—Odoo's intention to evolve from an "SME ERP" to a "global AI management platform" is very clear.
These three things point in the same direction: ERP is evolving from a "system of record" to a "system that can work on its own." But how to change and at what pace—different vendors give different answers. I have organized them into three routes as a reference for enterprise managers who are currently making their selection.
Route One: Odoo's Open Source Path—AI Embedded in Everything, but Barriers Left for the Ecosystem
The core change in Odoo 20, to put it in one sentence, is that AI has shifted from being "attached" to being "embedded."
Looking back at Odoo 19, AI was more like a standalone application. You open an AI window, ask it a question, and it answers. By Odoo 19.3, AI began to have autonomous operation capabilities — reading PDFs to generate purchase orders, building websites using natural language. But essentially, it was still "you tell it to do something, and it does it."
Odoo 20 is different. I carefully reviewed the official roadmap and interpretations from various partners, and several key changes are worth noting:
Module | Traditional Way | Odoo 20 AI Approach |
Accounting | Manual entry, monthly reconciliation | AI automatic audit, anomaly detection, natural language summary |
Inventory | Replenishment based on safety stock | AI predicts demand, dynamically optimizes replenishment plans |
Working hours | Manually fill in work order | AI analysis of actual activities automatically suggests working hours |
CRM | Sales manually assign customers | AI monitoring pipeline, dynamic reallocation of leads |
Project | Manually set up project structure | AI reads PDF to automatically create projects, set phases, and establish tasks |
Mobile | Online operation required | Full offline support, enabling operation and query even without internet connection |
Once this capability is implemented, a large number of repetitive tasks in daily operations—such as invoice entry, inventory reconciliation, scheduling and production planning, and customer assignment—will be directly taken over by AI. People only need to review exceptions and make decisions.
There is a fundamental difference between Odoo's approach and that of large ERP vendors: it does not create a closed system. The native MCP server allows AI to directly read and write Odoo data through the /mcp endpoint, without needing to install an additional Python gateway. Moreover, even if you are not using the Enterprise version, there is a way to connect — the Community edition plus a self-configured MCP Server, with the open-source ecosystem filling the gap. This "open foundation + AI embedding" approach is Odoo's greatest differentiating advantage in the AI era.
Significance for business managers:
If your company is already using Odoo, or is evaluating ERP, Odoo 20 provides a low-cost path to gaining AI capabilities. It doesn't require you to do everything at once — you can start with the AI features of one module, run a pilot, and then expand. Moreover, because it's an open-source ecosystem, you have no risk of vendor lock-in.
Route 2: The "Self-Built OS" Path of Large Vendors — From Delivering Tools to Delivering Intelligence
Kingdee took a different path. Liu Zhongwen stated quite bluntly in an interview — "Lingji" is not about stacking features, but about rebuilding the underlying architecture for the era of intelligent agents.
Kingdee's approach is: future enterprise software should no longer consist of independent functional modules, but should instead be an "operating system" that AI can understand business, execute tasks, and be governed. Lingji breaks down AI capabilities into a six-layer architecture, from the model layer to the application layer and then to the ecosystem layer, attempting to solve the integration problem of AI and ERP with a unified underlying foundation.
This path is deep enough, but the cost is also significant. Liu Zhongwen himself admitted that this is not something a single company can accomplish independently. By opening up both the Skill market and the Agent market, Lingji is essentially bringing the ecosystem into the fold.
Yonyou hasn't been idle either. Changjietong's "Xiao Chang" just obtained the first batch of ERP intelligent agent capability certification from the China Academy of Information and Communications Technology, and YonSuite is also promoting AI agents to rewrite the procurement logic of SaaS ERP. However, the approach is not entirely the same as Kingdee's—Yonyou is more inclined to embed AI capabilities within existing SaaS products rather than starting from scratch to build a separate operating system.
"What enterprises need in the future is not more single-point AI tools, but a set of underlying operating systems that enable AI to understand business, execute business, and continuously grow."
——Kingdee Vice President Liu Zhongwen
Route 3: Huawei Cloud's Ecological Empowerment Route — Open Source + Cloud Native Builds AI Foundation
The third path comes from Huawei Cloud. In early July, Huawei Cloud partnered with Suda Software to launch the "Honor Series ERP," which comes in two versions: an open-source edition and a super user edition. The core idea is to directly embed Huawei Cloud's AI inference engine, Ascend computing power base, and cloud-native architecture into the ERP system.
This approach is pragmatic: instead of building ERP systems themselves, they provide AI infrastructure to those who do. The Honor series open-source version targets enterprises and software service providers with deep customization needs, while the super user version is aimed at enterprise customers who use it directly. Essentially, it uses Huawei's AI capabilities to "replace the core" of traditional ERP.
This path is more suitable for two types of enterprises: one is those already using Huawei Cloud and hoping to integrate their ERP with the cloud infrastructure; the other is those with strong data sovereignty needs—Huawei's solution has a natural advantage in data localization.
Three routes viewed together: three choices for business managers
The logic of the three routes is actually quite clear. Odoo follows an "open foundation + AI embedding" approach, emphasizing low barriers, scalability, and no lock-in. Kingdee/Yonyou adopts an "AI-native OS" approach, focusing on deep integration, underlying reconstruction, and ecosystem co-construction. Huawei Cloud takes an "infrastructure empowerment" route, emphasizing AI computing power, data sovereignty, and cloud-native transformation.
No route is absolutely correct. The key lies in where your business's starting point and needs are.
I have compiled three multiple-choice questions for you to answer based on the actual situation of your company:
First question: Is your ERP currently "usable" or "easy to use"?
If your current system is still addressing basic functional coverage—finance, procurement, inventory, and production are not yet running smoothly—then don't rush to implement AI. First, sort out your data and processes. The weekly report of Chilong ERP mentions a statistic: "Among ERP products on the market that claim to offer 'business-finance integration,' less than 30% truly achieve automatic generation of accounting vouchers from business documents." If the data flow isn't connected, deploying AI will be futile.
Second question: Do you want "AI capability" or "AI architecture"?
If your needs are very specific—such as wanting financial reconciliation to be faster or inventory predictions to be more accurate—then Odoo 20's module-level AI features can meet your requirements without waiting for some "AI operating system." However, if what you're after is the intelligent transformation of the entire organization in three to five years—where the ERP not only records business but actively participates in business execution—then you do need to consider OS-level capabilities. The former can show results in three months, while the latter may take two to three years.
Third question: How high are your data sovereignty requirements?
For overseas enterprises, regulated industries, and companies with strict data localization requirements — data sovereignty must be a top priority when selecting a solution. Odoo's open-source model allows for private deployment, keeping 100% of data in your own hands. Kingdee/Yonyou's cloud deployment follows domestic compliance pathways. Huawei Cloud + Suda can run within domestic Xinchuang environments. Choosing the wrong option will result in very high migration costs later.
This Thursday's Odoo Shenzhen event is a window for observation
On July 16, Odoo held the "Digital Intelligence Innovation Day" at The Langham Shenzhen, specifically for overseas enterprises to discuss how AI+ERP can be implemented. I took a look at the event agenda, which includes full-scenario practical demonstrations for going global, customer case sharing, partner practical experience sharing, and an interactive experience zone.
If you are in Shenzhen, or if your company is in the overseas business, it is worth a visit. Not to say that going there will give you a standard answer—rather, this type of event allows you to see at once: to what extent Odoo has implemented AI, what level your peers' practices are at, and how well the ecosystem can support AI capabilities. This information is more useful than reading ten vendor whitepapers.
Summary
In July 2026, the AI transformation of ERP reached a clear boundary.
The China Academy of Information and Communications Technology has set standards for AI capabilities, and Chanjet has obtained the first batch of certifications. Odoo 20 has embedded AI into every module. Kingdee has built an AI operating system from scratch. Huawei Cloud has given ERP an AI core. The management complexity faced by overseas enterprises is rapidly increasing.
At this point in time, the most valuable thing for business managers is not to worry about "whether I am falling behind," but to figure out two things:
First, what is the most painful link in your enterprise right now—is it slow financial reconciliation, inaccurate inventory, failure to track sales leads, or inability to manage cross-border operations—starting from the most painful point is where AI can deliver results.
Second, choose a partner that allows you to "progressively upgrade." Whether you choose Odoo, Kingdee, or Huawei Cloud—the key is whether this system can let you start with a small scenario and gradually expand to the entire organization. Those who jump straight into "full AI transformation" will likely end up dizzying themselves first.
The era of ERP "doing its own work" has arrived. The pace is not the key; as long as the direction is right, it will do.
