In brief An AI generates code in a few minutes, but an ERP carries ten years of accumulated experience, regulatory compliance and edge cases that nobody had ever documented. AI and ERP are not in opposition: connected to controlled data, AI accelerates the ERP. Separated from it, it produces plausible prose without guaranteeing either the year-end close or the audit.
On LinkedIn and at conferences, the message is spreading: AI can now replace an ERP in a matter of days. A few natural-language queries, a quote appears, an invoice is generated, the room applauds. We are still waiting for the ERP and AI demo that survives a year-end close.
A demo is not a system
On a screen, a few minutes are enough to convince. The technique is nothing new: ERP vendors have been doing it for thirty years and it still works just as well.
In a small or medium-sized business, reality is less photogenic. The system must survive stocktaking, the statutory auditor's review and the tax audit that calls for documents from three financial years. It must also survive the departure of the one person who understood the product structure, the one whose knowledge nobody thought to document.
Between a demo that impresses and a system that holds for ten years, there is a world of difference. That world has a name: live operation.
What a management system really contains
An ERP is not just a programme. It is first and foremost a body of accumulated experience, built up version after version across tens of thousands of clients.
Take accounting: posting rules, exchange differences, reversal entries, margin VAT schemes, reverse charge. Each case was discovered, corrected and consolidated over the years, often as a result of a client in difficulty. Take logistics: batch and serial numbers, forward and backward traceability, multi-level bills of materials, partial receipts. These edge cases were documented nowhere: they lived in the head of the accountant or the warehouse manager.
An AI writes code in a few minutes, that is a fact. But that code does not know what it does not know. It is unaware of the customer who delivers in two shipments, with a deposit and a partial credit note. It will only find out at the close and on that day, the error falls on the business owner.
Compliance cannot be generated
The vendor carries a second legacy, less visible: regulatory obligation. The accounting entries file, electronic invoicing, VAT, document retention. These rules evolve from one year to the next, sometimes from one quarter to the next and it is the vendor who monitors them on behalf of all its clients.
Replacing the ERP with a generated assembly means taking on that monitoring permanently. Who will track changes to the FEC format? Who will adapt the data flows to the approved platform? Who will answer the auditor when asked why an entry was modified after validation?
A statutory auditor is not satisfied with a correct result alone. They want to know how that result was produced, by what proven system and according to what auditable rules. An AI writes code. It carries neither compliance nor an auditor's trust.
AI accelerates those who know what they are looking for
Make no mistake: at ANOR, we use AI every day and in highly advanced ways. Requirements analysis, code generation, documentation, acceptance testing: it plays a part at every stage of our projects. We cannot be accused of being lukewarm about it.
It is precisely because we practise it that we understand its limits. It considerably speeds up the work of someone who knows what they are looking for. It brings a great deal to the expert and far less to the person who hopes to do without one. Code produced in a few minutes is still code that needs reviewing, testing and maintaining and reviewing it requires knowing the client's business as well as the technical framework.
Odoo 20 integrates AI in a particularly advanced way: agents in sales, purchasing, inventory, helpdesk and project management, AI-calculated fields and connection to MCP protocols. This foundation rests on PostgreSQL, a free and high-performance database and on Python, the dominant AI language today. AI is connected here to real, controlled and traceable data.
AI without an ERP chatters; an ERP without AI works more slowly
AI only delivers its full value once connected to company data: quotes, stock, invoices, accounting entries. Without them, it produces plausible prose. With them, it becomes useful.
- It chases the customer whose payment is overdue.
- It flags a stock shortage before it happens.
- It reconciles a bank statement.
- It prepares a response to a request for quotation drawing on the real price history.
Conversely, an ERP without AI fulfils its role, but more slowly than it could. Repetitive data entry, tedious searches, letters to draft: that is where time is saved. Combining the two saves time and that is the whole point of the pairing. Our organisation and process management consultancy always begins by identifying where that time is being lost, before introducing any tool at all.
Five questions to ask before being won over
When faced with a promise of replacement by a simple query, an SME owner would do well to ask a few basic questions.
- What happens at the year-end close? Who guarantees the accuracy of the entries?
- How does the system respond to a tax audit covering three financial years?
- Who handles regulatory updates when the law changes and at what cost?
- What happens to the business if the provider disappears or the tool's designer leaves?
- Where are the references from comparable businesses, after several years of real live operation?
If the answer fits into a ten-minute demo, the question has not been understood.
Our position: AI in the ERP, with method
We connect AI to Odoo with method: on controlled data, within a framework that preserves traceability, with users who remain in control. That is where the real gains lie. To assess your current situation, our complimentary situation audit helps identify priorities with complete clarity.
As for replacing an ERP with a simple query, we leave that to science fiction. Speak with a consultant to take stock of your project, with no obligation.
Frequently asked questions
Can AI really replace an ERP in a small or medium-sized business?
No, not in the current state of things. An ERP carries years of accounting rules, regulatory compliance and complex business cases. An AI generates code quickly, but that code is unaware of situations it has not yet encountered and it is often the year-end close or the tax audit that reveals them.
How is AI used in Odoo 20?
Odoo 20 integrates AI in a particularly advanced way: agents in sales, purchasing, inventory, helpdesk and project management, AI-calculated fields and connection to MCP protocols. AI acts on real and traceable data, which makes it genuinely useful day to day.
Who handles regulatory updates if I use Odoo?
It is the Odoo vendor who monitors regulatory changes: the FEC format, electronic invoicing, VAT and document retention. Each update is rolled out to all clients, without each business having to manage that monitoring internally.
Can AI speed up my use of Odoo?
Yes, that is precisely where it adds the most value. Connected to your Odoo data, it chases overdue customers, flags stock shortages, reconciles bank statements and prepares responses to requests for quotation. Without that data, it is nothing more than a plausible text generator.
How can I tell whether my current system is robust enough to last ten years?
ANOR offers a complimentary situation audit to assess your organisation and your tools. A thirty-minute conversation is often enough to identify weaknesses and priorities, before any discussion of tools or migration.
