McKinsey says ERP is being ended by AI — implementation costs halved, opportunity for open-source ERP arrives

In May of this year, McKinsey published an article with a title so straightforward it made people take a second look: "The End of ERP as We Know It?"

The major clients that had served this company for thirty years—SAP, Oracle, and the like—suddenly came forward and said, "The ERP you're familiar with might soon be gone." This signal was more striking than any data in the report.

I carefully read the original text three times. The conclusion is: McKinsey is not playing clickbait. Each of the five disruption paths it outlines is redefining the ERP category that has existed for over forty years. What concerns me most is not how SAP or Oracle will respond—it's that open-source ERP like Odoo has secured a very special position in this round of reshuffling.

The most explosive data: implementation cost and time both cut in half

Let's first look at a set of numbers.

McKinsey clearly stated that AI agents can reduce the workload of ERP implementation by at least 50%, and shorten the implementation cycle by half as well. You read that right, it's not 10% or 20%, it's 50%.

Let's break down the specific changes in capability:

• Target Process Design: AI agents complete process design recommendations in days that previously took months, with extremely high accuracy
• Automatic configuration: The recommended solution is directly implemented into the ERP system, with humans only performing quality control.
• Data Migration: The traditional cleaning process that took months has been significantly compressed by AI
• Testing and Documentation: Heavy automation in test case generation, documentation writing, and training preparation

However, McKinsey also issued a warning — for every 1 yuan spent on developing an AI model, 3 yuan must be allocated for change management. The interesting part is that the implementation costs cut by AI are largely the consulting fees of existing vendors, while the money for change management cannot be saved.

Two Futures: ERP Completely Replaced, or Becoming "Invisible"

McKinsey describes two extreme scenarios.

Radical (SaaSpocalypse)

ERP completely disappears. AI agents dynamically create and optimize processes. Data resides in microservices, and application logic becomes AI-driven. Screen operations are replaced by natural language interaction. Users speak to the system, and the system decides what to do on its own.

Gradualists (Stable Core)

ERP remains the backbone infrastructure—the data layer, application layer, and front-end skeleton continue to exist, ensuring compliance, auditing, and scalability. However, users no longer interact directly with ERP; AI has become the intermediate execution layer.

The two scenarios are not mutually exclusive. McKinsey itself states in its report that the most likely outcome is a fusion of both—ERP will not disappear, but users will become "invisible" to it. AI becomes the new interface for human-computer interaction, where users express intent, verify results, and intervene in anomalies, while all intermediate execution is completed by AI.

Translation into plain language: In the future, your finance staff may not even need to open the finance module interface. They just need to tell the AI, "Follow up on last month's receivables," and the system will handle it on its own. The ERP runs in the background, while people interact with AI in the foreground.

Looking back at Odoo 20's roadmap, you'll understand why I particularly care about this conclusion. Odoo 20's native MCP server allows AI to directly read/write Odoo data, with Agentic AI handling financial audits, inventory forecasting, and lead allocation. It is transforming itself into AI's data foundation, not an operation interface—exactly the path McKinsey described.

Gartner says 62% of funds flow to AI-native

McKinsey is not the only one thinking this. A set of data from Gartner makes this shift more tangible:

62% of cloud ERP spending in 2026 will go to AI-native platforms (only 14% in 2024), a 48 percentage point share shift within 24 months.

What does 48 percentage points mean? The closest historical reference is the SaaS migration in the late 2000s, which took a full decade to reach a similar tipping point in market share. The migration to AI-native ERP has compressed this timeline to one-fourth of that.

The AI-in-ERP market size has reached $7.33 billion in 2026. Precedence Research predicts it will grow to $58.7 billion by 2035, with a compound annual growth rate of 26%. Moreover, this figure may be conservative—McKinsey and Gartner are both saying the same thing; when viewed together, the direction is clearer than any individual forecast.

Structural advantages of the Odoo open-source model in this round of reshuffling

I discussed this with a few friends who implement ERP systems, and the general consensus is that Odoo's structural advantages in the AI era may be greater than imagined.

I roughly summarized:

First, Odoo's modular architecture is naturally suited for "headless ERP." McKinsey's notion of "ERP becoming the backend" presupposes that the ERP itself can be broken apart. The integrated architectures of SAP and Oracle are not impossible to dismantle, but doing so comes with extremely high costs and significant risks. Odoo's modularity was designed from the outset for disassembly and recombination—each module can be independently upgraded and independently integrated with AI. Its native MCP server embeds the AI interface directly into the product, not as a plugin or gateway, but as part of the kernel.

Second, Odoo's implementation cost itself is "halved." Traditional ERP implementation often costs hundreds of millions of dollars, and even after being halved, it remains an astronomical figure. How much does Odoo Enterprise cost annually? Small and medium-sized enterprises can afford it. When the dividends of AI-driven cost reduction and efficiency improvement are primarily realized by SMEs, Odoo's user base (2.11 million paid users, 67% served by partners) provides a vast pool of implementation scenarios. McKinsey's claim that "change management costs double" is also easier to execute in SME scenarios—fewer people, lower communication costs.

Third, the AI integration speed of the open-source community is faster than that of closed-source vendors. Odoo 20's native MCP server was built directly into the core by the R&D team. In the community, people are already developing AI Agent templates based on MCP—procurement agents, inventory forecasting agents, and financial audit agents. Once this ecosystem takes off, Odoo will be much faster than SAP/Oracle in filling industry gaps.

"Odoo becomes the MCP server itself, without any added layers." —— Odoo R&D Team

What should Odoo users do now

The McKinsey report has been out for two months, and Odoo 20 is set to be released in September. I’ll pick three of the most practical suggestions:

First, don't wait until version 20 is released to study the upgrade path.Odoo 20 has many breaking changes — the XML-RPC API needs to be migrated, the permission model is redesigned, and custom modules require compatibility testing. Now is the time to start sorting out your customization list, listing dependencies, and performing upgrade tests in a sandbox. If you don't prepare in advance and panic after seeing the release news in September, it will be too late.

Second, start with small-scale AI integration. First, use Odoo 19's built-in AI features (OCR invoice recognition, CRM lead scoring) to get the team accustomed to a "conversational" workflow with the system. By the time version 20 is released, the native MCP server will be directly available, transforming AI from an assistant into an executor, and the team won't find it too abrupt.

Third, treat change management as a serious matter. McKinsey says "for every $1 spent on the model, allocate $3 to change management" — this is no empty talk. The implementation costs cut by AI are actually redistributed to organizational transformation. If Odoo implementation partners can help clients redesign processes and train personnel, merely being "50% cheaper than others" is no longer enough — they need to make clients feel that "by switching to Odoo, AI has also helped me manage things well."

At the end of writing, I want to say one thing. I have read this McKinsey article over and over again several times. What touched me the most was not the numbers, but the fact that this company is willing to publicly say, "Your way of playing might be about to become obsolete," based on the game rules it established itself. Thirty years ago, it helped SAP overthrow the previous generation of ERP, and today it is paving the way for the new generation of ERP.

This time, it's the turn of open-source players like Odoo to sit at the table.