If you follow open-source ERP, you may have noticed something recently when opening the Odoo official website: the registration page for Odoo Experience 2026 is already live, from September 24 to 26, at Brussels Expo, with the Basic Pass free of charge. The opening keynote speaker is founder Fabien Pinckaers, who gets straight to the point right from the start—releasing Odoo 20.
This is not an ordinary annual update. Over the past three years, Odoo's AI path has progressed from "content generation" (v17) to "practical tools" (v18), and then to "AI infrastructure" (v19, with Agent, Ask Odoo, and voice commands). The goal for v20 is very clear: shifting from "you ask, it answers" to "it completes an entire workflow on its own." In the words of the official preview, this is called Agentic AI—agentic intelligence, where the system no longer waits for you to click buttons, but instead proactively monitors processes, executes tasks across modules, and directly handles issues as they arise.
In this article, I want to combine the conference information from Odoo Experience 2026, the pre-release interpretations from multiple partners, and the list of features already disclosed for Odoo 20 to discuss three things: how significant this release really is, which specific changes businesses should pay attention to, and what preparations should be made in the next two months.
I. From "Q&A" to "Execution," Odoo's Three-Year Leap in AI
To understand why Odoo 20 is called "the most transformative release in the industry's history," first look at its AI evolution path over the past few years, which is actually a very clear straight line:
| 版本 | Time | AI Capability Positioning | Typical capabilities |
|---|---|---|---|
| Odoo 17 | October 2023 | 内容生成 | Marketing copy, website text, basic AI suggestions |
| Odoo 18 | October 2024 | Utilities | OCR enhancement, automatic invoice processing, data cleaning |
| Odoo 19 | October 2025 | AI Infrastructure | AI Agent, "Ask Odoo" natural language search, voice-to-command |
| Odoo 20 | September 2026 | Agentic AI autonomous execution | Autonomous completion of multi-step workflows, cross-module collaboration, and proactive alerts |
Note the difference between v19 and v20 — the former, despite being called an Agent, is still essentially "reactive": you ask, it answers; you tell it to do one thing, it does one thing. The Agent in v20, however, is "proactive": for the ticket you're currently working on, it has already automatically summarized the historical conversations and drafted a reply; for leads that have been sitting in the CRM for a long time without follow-up, it scans them and directly reassigns them according to the rules; for the timesheet on Friday afternoon, it has already auto-filled most of it based on your emails and submission records.
In plain terms: v19's AI is like a new employee on standby—say something and it acts; v20's AI is like a veteran employee with three years of experience—no need for instructions, it knows what to do next.
Based on the teaser information released by partners, Odoo 20's agentic capabilities have already been implemented across multiple modules: Helpdesk's AI copilot automatically summarizes lengthy ticket conversations and drafts replies based on historical solutions; the project module's AI can read technical design documents (PDFs) and automatically generate a project structure with phases and tasks; timesheets feature a privacy-friendly background tracker that suggests time entries based on your activities such as pull requests and emails. These are not just concepts on a slide deck, but features explicitly listed in the release notes.
Second, besides AI, there are three major things in this version
Agentic AI is the face of this release, but the three other things may have a greater impact on enterprise technology selection.
1. Read-Replica Database Architecture: Breaking Through the 10,000 Concurrency Ceiling
This is the first time Odoo has officially positioned itself at the architectural level for "large enterprises." In the past, Odoo defaulted to a single database handling all reads and writes; v20 splits responsibilities: the primary database handles transactional writes, while read-replicas handle dashboards, reports, and analytical queries. The official goal is straightforward—supporting 10,000+ concurrent users. Previously, many enterprises hesitated to use Odoo for group-level deployments, with the core concern being concurrency; this architectural change effectively pulls Odoo into the selection list that previously belonged to SAP and Oracle.
2. Financial closed-loop capability significantly strengthened
The accounting module in v20 does several very practical things: Pay from Odoo — pay directly within the ERP via SEPA or single-signature bank batches, no need to switch to online banking; AI Accounting Audit — automatically runs analytical reviews, identifies anomalies, verifies cash inventory, and outputs conclusions in plain language; Reconciliation for Any Account — any account that triggers reconciliation will automatically generate adjustment entries such as revenue deferrals and exchange differences. For finance teams, these are things used every day, not just icing on the cake.
3. New Industry Verticals and Channel Expansion
Hotels (check-in management, seasonal rate tables), construction (WBS planning, subcontracting, progress billing), and recruitment agencies (CRM-recruitment-invoicing integration) are new industry packages added in v20; on the retail side, native connectors for TikTok Shop, Amazon, and Lazada have been added, and POS hardware configuration has become much simpler thanks to Chromium's local network access capability. Enterprises engaged in cross-border and chain retail should pay special attention to this part.
3. What should enterprises do in the remaining two months?
The release cadence for Odoo 20 is actually already clear: the keynote kicks off on September 24, with the stable version expected to be released between October and November. For existing users and businesses planning to get on board, my advice is not to wait until the day of the release to take action.
Four things you can do right now:
- Schedule the "upgrade assessment": Put the Odoo 20 upgrade on the roadmap in Q3, assess the compatibility of current custom modules and third-party applications with the new version, and leave buffer time for the development team to make adjustments.
- First, clean up data quality: Multiple partners have repeatedly emphasized in their warm-up articles that data quality is a prerequisite for the implementation of Agentic AI. The premise for AI to execute automatically is that the data is trustworthy; dirty data will only cause AI to amplify errors. Taking advantage of the upgrade window to do a round of master data cleaning is much more cost-effective than rework later.
- Pick 2-3 high-value scenarios for pilot testing: Don't roll out all modules at once. Low-risk, high-frequency scenarios like automatic summarization of customer service tickets, automatic lead reassignment, and automatic timesheet filling are best suited for initial implementation, then gradually expand to procurement and finance.
- Follow the official release pace: The September 24 Keynote will feature a full functionality demo, with a live stream on YouTube; tickets refunded before August 15 can still get 50% back, and companies with budget and visa conditions can consider attending in person—after all, three days can pack in 400+ sessions, which is far more efficient than reading secondhand accounts.
One point that needs to be flagged in advance: Agentic AI means systems start "taking initiative," so permission models and approval boundaries must be tightened in tandem. Odoo 19.4 has already refactored the permission system into a unified ir.access model, and v20's agentic AI will inevitably rely heavily on this permission framework—enterprises must treat "what AI can and cannot do" as a mandatory question during upgrades, rather than discovering it only after going live.
