Introduction: ERP sellers have started selling "implementation" itself
In the past, selling ERP meant selling licenses and earning from implementation fees. Requirements research, blueprint design, configuration, testing, data migration—every step was piled up by consultants billing by person-day. This tacit understanding was directly punctured by Oracle this month.
A November 3 analysis by the person known as T客 hit the nail on the head: at the end of October, Oracle played two cards—one on product and one on implementation—and the implementation card is a fundamental blow to the traditional ERP delivery model.
1. Oracle's Two Cards: Prebuilt Skills + AI Implementation
The first is SuiteSkills: 45 preset AI skills covering finance, inventory, projects, and sales. The key is that it is not tied to its own large model; customers can connect third-party models, but permissions and role controls remain locked within NetSuite. This approach is very smart—capabilities are opened to customers, while governance stays with the vendor.
The second one is even more aggressive: SuiteExpress, which uses AI directly for implementation. It captures requirements, maps processes, migrates data, and validates workflows. The official line is that it compresses the implementation cycle from "several months" down to "several weeks or even days." NetSuite carries 28 years of implementation experience from over 40,000 customers, and the AI uses this historical data to match new customer requirements, automatically generating configurations and running tests, reducing the consultant's role to only making "key decisions" and "handling exceptions."
The article's judgment hits the nail on the head: what Oracle is cutting is not ERP, but the per-person-day pricing model of ERP implementation. The consultants in the middle layer who rely on information asymmetry and template reuse are the most at risk.
II. CNNC Enterprise Intelligence Hub: Independent and Controllable Becomes a Hard Constraint for Large Groups
Turning our attention back to China. On October 23, China National Nuclear Corporation (CNNC) released “Qizhishu ERP,” a new-generation intelligent hub for the operations and management of central and state-owned enterprises. Built by Tongfang Co., Ltd. (CNNC Nuclear Information), it is the Group’s “No. 1 Project” for digital transformation. It adopts a full-stack solution of “one core ERP + N scenario-integrated applications” and has been deployed at more than 1,400 legal entities of CNNC at home and abroad. It adheres to a full-stack, independently controllable, cloud-native, composable and componentized approach, and is deeply integrated with large AI models, driving process-oriented ERP to leap toward Agentic intelligent agents capable of perception, early warning, and analysis and judgment.
The significance of this is not that “another ERP has been released,” but that it turns “independent and controllable” from a slogan into a hard constraint in the digitalization budgets of large central and state-owned enterprises. When data ownership and a controllable foundation become procurement prerequisites, the weighting table for ERP selection is rewritten.
3. Open-source alternatives gain momentum: Odoo and ERPNext fill Sage's gaps
This "sovereignty + cost" consideration is spilling over to mid-sized enterprises. In its 2027 open-source Sage alternatives directory updated on November 1, Fe/male Switch lists ERPNext (GPLv3), Dolibarr, Tryton, iDempiere, and Odoo Community (LGPLv3) together as full ERP options to replace Sage 200 / Intacct / X3.
But the catalog also honestly reminded us: although Odoo Community is open source, the complete Accounting application is proprietary to Enterprise. This is precisely the pitfall most easily encountered when selecting open-source ERP—being able to see the source code does not mean being able to use all the features. For enterprises that want to use open source to gain sovereignty and cost advantages, first clearly understand the license red lines, then calculate the total cost of ownership; this is far more important than blind migration.
IV. It's Time to Change the Selection Criteria
Putting these three things together, the logic behind ERP selection has actually quietly changed its yardstick.
Previously, the comparison was over whose feature list was longer and whose demo was prettier; now three questions must be asked: whether sovereignty is independent and controllable, whether results are verifiable, and who has the final say on cost. Oracle turning implementation into a product means that the vague cost item of 'person-days' will be amortized into subscriptions; CNNC's independent and controllable practices mean that the say over the foundation layer cannot be lost; and the warming of the open-source camp gives enterprises a realistic path of 'owning their data and not being locked in'.
For growth-stage companies, my advice is: don't get thrown off course by the 'AI-native' narrative. First connect your existing ERP to AI through connectors and automate the most painful processes—this is more reliable than tearing everything down and starting over. At the same time, free implementation costs from the 'daily billing' mindset and focus more on 'output per unit process'.
Conclusion: Delivery shifts from stacking person-days to AI plus fewer consultants
Oracle takes implementation, China Nuclear guards sovereignty, and open source grabs replacement—these three forces point to the same inflection point: ERP delivery is shifting from 'stacking person-days' to 'AI plus a small number of consultants,' while selection is shifting from 'comparing checklists' to 'comparing sovereignty, comparing results, and comparing cost controllability.'
For implementation vendors, this is a warning that their moat is being filled in; for enterprises, it is a window to obtain better digitalization at a lower cost. During this window, whoever first figures out that what they want is “controllable results” rather than “more features” will be the first to reach shore.
