Short answer: AI coding tools let you write Odoo modules fast and cheap, but the risks of AI-generated code are real: in Veracode's 2025 study, language models introduced serious security flaws in 45% of test tasks. AI is useful for drafts and documentation, but accounting logic, access rights and version upgrades need review by an experienced specialist. Deep Vision uses AI tools, but responsibility for every line of custom code stays with certified Odoo experts.
Today you can hand a spec for an Odoo module to a chatbot and get working code in minutes. Some providers sell this as “fast and cheap.” For a business owner the question is simple: will this code break my reports, is my data safe, and what happens when we move to the next Odoo version? This article looks at the risks of AI-generated code using public research and separates where AI helps from where expert review is not optional.
What is AI-generated Odoo code?
AI-generated Odoo code is a module, or part of one, in Python, XML and JavaScript produced by ChatGPT, Copilot or another language model.
An Odoo module is more than code. It includes a data model, views, access rights, business rules and often accounting entries. A language model can write all of that with correct syntax. But it does not know your business, your accounting policy or the roles in your company. The result is code that runs, but runs wrong.
What does research say about the risks of AI coding?
Two recent public studies put clear numbers on this.
Security. In 2025 Veracode tested more than 100 language models on 80 coding tasks. In 45% of test cases the models introduced vulnerabilities from the OWASP Top 10 list, and for Java the failure rate was above 70% (Veracode). The OWASP Top 10 is the standard list of the most common web risks, such as SQL injection and broken access control. Put simply, in nearly half of cases, unreviewed AI code carries risk.
Speed. In July 2025 the research group METR followed 16 experienced developers through 246 real tasks. With AI tools they took 19% longer. Before starting they expected a 24% speed-up, and afterwards they still believed they had been about 20% faster (METR). AI creates a feeling of speed, not always real speed.
None of this says AI is bad. It shows one thing: you cannot hand responsibility for the result to AI.
Where is AI genuinely useful in an Odoo project?
Deep Vision is not against AI tools. Used well, they save time:
- Drafts of simple code and repetitive templates, such as XML for a report view.
- First versions of technical documentation, comments and user guides.
- Building a list of test scenarios.
- Reading error messages and searching for likely causes.
- Helper scripts for data import, always reviewed afterwards.
What these tasks share: if AI gets something wrong, you see it quickly and it does not hurt the business.
Where is expert review mandatory?
In some areas an AI mistake is invisible at first and costs a lot months later.
Accounting logic. Costing, inventory valuation, VAT and journal entries. AI will write neat code that may post a debit or credit to the wrong account. That kind of error appears at year end, when you file reports. More on the module: Odoo accounting module.
Access rights. In Odoo, who sees and changes what is controlled by access rights and record rules. AI often adds a sudo() call or widens permissions “to make it work.” The result: a sales manager can see cost prices or another branch's data.
Version upgrades. Odoo ships a new version every year and internal APIs change. Language models are often trained on examples from older versions and suggest deprecated methods. That code works today and breaks at the next upgrade.
Integrations. Connections to Didox, Asl Belgisi, banks and payment systems involve secret keys, error handling and legal requirements. Unreviewed code here is a direct financial and legal risk.
Performance. AI code runs fast on a small test database and slows down with 7,000 products and hundreds of users.
How can you check how a provider uses AI?
As the client, ask these questions and write down the answers.
- Who reviews the code, and in what order (code review)?
- Are automated tests written for custom modules?
- Are access rights and record rules checked separately?
- Which Odoo version is the code written for, and is there a plan for the next one?
- Will you receive the source code, and is it kept in Git?
- Does the team include Odoo-certified experts (check in the odoo.com directory)?
- Is a standard Odoo feature considered first, or is new code written straight away?
- Who fixes a bug during the warranty period, and how fast?
If the answer sounds like “AI writes it, we deploy it,” treat that as a warning sign. The price looks low, but the risk stays with you. On custom builds versus ready systems: build an ERP from scratch or use a ready one.
Why does cheap AI code end up expensive?
Writing code is only part of a project's cost. The main costs come later: finding bugs, fixing data, rewriting code at each version upgrade. If nobody deeply understands a module's logic, every change creates a new problem. That is why experienced teams keep custom code to a minimum and use standard Odoo as much as possible. On the editions: Odoo Community vs Enterprise.
How does Deep Vision use AI?
Deep Vision is an official Odoo Gold Partner from Tashkent. Its Odoo partner directory page lists 7 Odoo-certified experts (versions 17 to 19), 20+ client references and 89% customer retention. For Gold, Odoo requires 300 new Enterprise users a year, 6 certified staff and at least 80% retention (odoo.com).
Our approach:
- AI tools are used as an assistant for drafts, documentation and tests.
- Every piece of custom code is reviewed by a certified expert who is accountable for it.
- A standard Odoo feature is proposed first; code is written only when there is a real need.
- Accounting and finance settings are checked by experienced specialists who know finance.
- For Didox and Asl Belgisi we use ready, tested modules instead of writing code from scratch each time.
If you want the custom code in your current Odoo reviewed, or you are starting a new project, fill in the brief or see our Odoo implementation service. A guide to choosing a provider: how to choose an Odoo partner in Uzbekistan.
Frequently asked questions
Should AI-generated code be avoided in Odoo altogether?
No. AI is useful for drafts and documentation. What matters is that an experienced specialist reviews the code and takes responsibility for it.
What is the biggest risk of AI-generated code?
Invisible errors: wrong accounting entries, over-broad access rights and security flaws. Veracode found vulnerabilities in 45% of test cases.
Does AI make developers faster?
Not always. In the METR study, experienced developers were 19% slower with AI while believing they were faster.
How can I check the custom modules I already have?
Check that you have the source code and tests, review access rights and confirm which Odoo version each module targets. Deep Vision can run this audit for you.
Does Deep Vision use AI?
Yes, as a supporting tool. Responsibility for every custom module stays with certified Odoo experts.
Sources
- Veracode: 2025 GenAI Code Security Report
- METR: impact of early-2025 AI tools on experienced developers
- Odoo: partner tier requirements
- Odoo partner directory: Deep Vision
Keep your Odoo code safe and maintainable: fill in the brief or call +998 77 093 00 07.