Odoo 20 was unveiled on 24 September 2026 at Odoo Experience in Brussels, and it follows the pattern of the last several releases: one major version per year, changes across more than 50 apps, and a support clock that starts ticking the moment it ships. If you run Odoo today, the real question is not whether you will move to version 20 — staying on an unsupported version is rarely a strategy — but when. Upgrade in the first weeks and you get the new features early, along with first-release rough edges. Wait, and you trade those features for a calmer migration. This guide works through the decision properly: what pushes you toward upgrading now, what justifies waiting, and how to match the answer to the version and setup you actually have.
The case for upgrading now
Your current version may already be out of support
The strongest argument has nothing to do with features. With Odoo 20 released, the supported versions are 18, 19, and 20 — and standard support for Odoo 17 ended in September 2026. If you are on 17 or anything older, you are running without the safety net of official fixes and, more importantly for planning, each year you wait adds another version jump to the eventual migration. Upgrading from 17 to 20 is a bigger project than upgrading from 19 to 20, and it only gets bigger. For a database on 17, “now” is the right answer in almost every case: plan the move, run a test upgrade, and get back inside the support window.
The headline features solve real problems
Odoo 20 is not a cosmetic release. Three changes stand out for operational teams. First, AI agents: agents can create and update records — including from an uploaded PDF or file — answer questions about a file, generate images, and take voice-dictated instructions. They can be triggered by automated actions and scheduled actions, their runs are reviewable in the automation chat, and what were called agent “topics” are now called “skills.” Second, offline mode: users can create, edit, archive, and delete records and rerun earlier searches without a connection, with changes syncing when they are back online — aimed squarely at warehouse floors, construction sites, and field technicians. Third, accounting gets stricter and smarter: under the new bank consistency rule, every journal entry affecting a bank account must originate from a bank transaction, manual entries on bank accounts are no longer allowed, and there is a new reconciliation flow with a reconciliation summary report. Add a native MCP (Model Context Protocol) connector for tying an external AI assistant into your database within users' existing access rights, and there is genuine capability here, not just polish.
Licensing timing is worth a look
The Standard plan price is unchanged with Odoo 20, but the Custom plan rises by roughly 20% from 2027 — the first Odoo price rise in four years — and a new Light User licence (from EUR 7.90 per user per month in Europe) covers limited-access users such as employees who only need HR self-service. Regional pricing varies, so check odoo.com/pricing for your figures rather than relying on anyone else's. The upgrade decision and the licensing decision are separate, but reviewing them together is sensible: if a move to 20 coincides with a renewal or a plan change, model the 2027 Custom pricing before you sign anything.
The case for waiting
Customisations and integrations are the real cost
An upgrade is cheap for Odoo and expensive for everything attached to it. The XML-RPC API is deprecated — its db service is removed in 20 and the remaining services follow in Odoo 22 — so external integrations need a planned move to the JSON-2 API rather than an emergency rewrite. The frontend moves to OWL 3, and custom views or widgets may need rework. Access rights have been simplified in a way that matters to anyone with custom security: record rules are replaced by a domain defined on each access right, merging access control lists and record rules into one model — custom record rules need to be reviewed and re-expressed, not just carried over. If your database leans on custom modules, Studio changes, or third-party apps, the honest timeline is set by testing and fixing those, not by Odoo's release date. Waiting until your customisations are confirmed working on 20 is not procrastination; it is the project.
Workflow changes need training, not just migration
Some changes alter how people work on day one. Accounting teams must absorb the bank consistency rule and renamed payment statuses — “In Process” is now “Paid,” and “Paid” is now “Reconciled.” Approved expenses now generate draft bills instead of flowing through the removed standalone Employee Expenses menu item. Anyone using Field Service will find the standalone app discontinued, with its features — live technician map, route planning, travel fees, worksheets — now living in the Planning app. None of this is a reason to avoid 20, but all of it argues for upgrading when you have time to retrain, not the week before your busiest period.
AI is not free, and not urgent
If AI agents are your main reason to move, budget first: all AI features in Odoo 20 require IAP credits, and no plan includes them. The cost is per agent and per use, not per user. That does not make agents bad value — it makes them a line item to estimate with real workflows, rather than a reason to rush a whole ERP migration in October.
The decision matrix
| Your situation | Recommendation | Why |
|---|---|---|
| Running Odoo 17 or older | Upgrade now — start the project immediately | Standard support for 17 ended in September 2026; every year of delay adds another version to the jump. |
| On 18 or 19, few or no custom modules | Upgrade soon, on your schedule | The migration is mostly testing and training; bank consistency, offline mode, and AI agents are available as soon as you move. |
| Heavy custom modules or Studio customisations | Wait — test first, upgrade when fixes are proven | Your timeline is set by your code, not Odoo's release. Run a test upgrade early to size the work. |
| Integrations using XML-RPC | Wait until integrations are migrated | XML-RPC is deprecated in Odoo 20 (db service removed now; the rest goes in Odoo 22); systems using the db service break on day one, everything else needs a migration plan. |
| Custom record rules and complex access rights | Upgrade after a security review | Record rules are replaced by domains on access rights; permissions must be re-expressed and re-tested. |
| Peak season or year-end close is near | Wait until after it | Accounting workflows change (bank consistency, renamed payment statuses, expenses becoming draft bills). Do not learn them mid-close. |
| Field teams struggle with connectivity | Upgrade soon | Offline mode with sync-on-reconnect is built for warehouses, sites, and technicians — a concrete daily win. |
| Mainly excited about AI agents | Upgrade when ready; budget IAP credits first | Agents are genuinely capable, but every AI feature consumes purchased IAP credits that no plan includes. |
A middle path that usually wins
For most databases, the best answer is neither “today” nor “someday”: it is test now, produce later. Request a test upgrade through the Odoo upgrade platform, restore the result somewhere safe, and let your team work through their real processes — raising invoices, reconciling the bank, running the warehouse — while production stays untouched on your current version. Odoo provides a free Odoo 20 test environment for evaluating the new version generally (test databases are deleted weekly, so treat it as a sandbox, not your migration rehearsal). The test upgrade tells you exactly what breaks, what needs retraining, and what can wait. Then pick a production date in a quiet window with full information instead of guessing.
The bottom line
Upgrade now if you are on 17 or older, if your customisation footprint is small, or if offline mode and the accounting changes solve problems you feel weekly. Wait if custom code, XML-RPC integrations, or a looming peak period would turn the migration into a fire drill — but wait with a test upgrade running and a date on the calendar, not indefinitely. Odoo's yearly cadence means version 21 will arrive on schedule whether you are ready or not; the goal is to make 20 a planned project, not an emergency.
