The hardest part of an Odoo upgrade is rarely the technical migration. It's proving to the business that everything still works: every invoice type, every approval, every report finance relies on at month-end.
That proof comes from user acceptance testing (UAT). Writing good UAT scripts is slow, repetitive work, and it's exactly where AI saves real time, provided a person stays in charge.
Why UAT is the bottleneck in an upgrade
In a version upgrade (for example V17 to V19), the business expects "the same, but newer". So the testing has to cover what people already do, not just new features. That means:
- every standard process (quote to cash, procure to pay, record to report)
- every customisation and third-party module
- every entity, if you run several companies on one database
- the edge cases people only remember when they break
Written by hand, that is often hundreds of test steps.
Where AI helps
1. Turning process notes into test scenarios. Give AI a process description ("credit notes over ₹50,000 need finance-head approval") and ask for test scenarios: the normal case, the boundary (exactly ₹50,000), and the failure case (approval rejected).
2. Writing step-by-step scripts. From each scenario, AI can draft numbered steps with the expected result for each: menu path, fields to fill, button to click, what should appear.
3. Building the regression list from customisations. Paste the list of custom modules and their purpose, and AI can suggest which processes each one touches and therefore which tests must be re-run.
4. Keeping scripts consistent. AI is good at applying one template across hundreds of scripts: same columns, same wording, same sign-off fields.
Where AI must not be trusted on its own
- Menu paths and field names change between versions. AI may describe the old screen. Every script needs a dry run on the new version.
- AI doesn't know your business rules. It will invent a plausible approval limit if you don't give it the real one.
- Client data stays out. Use process descriptions and dummy values, not real customer or financial data, in any public AI tool.
A workflow that works
- Collect process descriptions and the customisation list from functional owners.
- Draft scenarios and scripts with AI, using a fixed template.
- Review each script with the functional owner: are the rules right?
- Dry-run a sample on the upgraded test database and correct menu paths.
- Execute UAT with business users, logging every failure as a defect.
- Sign off by process owner, with evidence attached.
AI can cut the drafting time sharply. The review and dry-run steps are what make the scripts trustworthy, and they can't be skipped.
A real example. On a recent Odoo version upgrade, our test plan covered dozens of test cases across six testing phases, each area mapped to a functional and a technical owner. For weeks, the test-coverage tracker barely moved. The reason was simple: some test areas had no named owner, so nobody started them. Once every area had one, testing moved.
Two more lessons from the same upgrade. Comparing the test plan with past incident tickets found risks the plan had missed. And the finance testers were in the middle of month-end close, so we started them on the processes that hadn't changed and saved the new ones for after close.
How we can help
RNXWe provide AI-assisted UAT and regression testing for Odoo upgrades, for companies directly or behind the scenes for Odoo partners.