Reviewing AI-generated X++ code: a 10-point checklist

AI-generated X++ code is reviewed like a developer's: through a pull request, with a fixed checklist. The ten points below follow the rules Microsoft documents for Dynamics 365 Finance & Operations. If one fails, the pull request goes back for rework: human approval keeps the final say.

1. Extensions, never over-layering: standard tables, forms and classes are changed through extensions. See Chain of Command vs event handlers.
2. Correct Chain of Command: next called at the first level, not inside an if or a loop; extension class marked final.
3. Build and BP check with no unjustified warning. See BP check warnings.
4. Security: every new menu item is covered by a privilege, linked to a duty and a role. See privilege, duty, role.
5. Performance: set-based operations rather than loops, joins rather than nested selects, no query in a form's init. See insert_recordset and update_recordset.
6. SysTest unit tests present, isolated (AutoRollback) and meaningful: they check the requested behaviour, not just that the code runs. See first SysTest test.
7. Naming: solution prefix, conceptual names, no underscores in data entities, Entity suffix.
8. Labels: all visible text in label files, no hard-coded strings.
9. Extensibility and dependencies: small well-named methods, no new delegates, minimal model references, no doInsert/doUpdate bypassing events.
10. Deployment: the package builds in the pipeline, database sync passes, and behaviour is checked on a test environment before production.

Without context, a model generates plausible but generic code: it ignores your prefixes, existing extensions, roles and data volumes. Points 4, 5, 7 and 9 are therefore the first to look at. Point 6 too: a test that always passes proves nothing. See what AI assistants can do in X++.

At Skalp AI, the agent generates the X++ code and its unit tests, compiles with a best-practice check, then pushes a pull request to the customer's Azure DevOps project with screenshots of the tests. Nothing reaches the working branch without the customer's technical architect reviewing and approving the pull request. See Azure DevOps integration, building X++ without a consultant in 2026 and the Basic, Pro and Max plans (1 credit, ≈ €400 on Basic, €200 on Pro, consumed only when the code is generated).

See also: AI ticket creation in D365 F&O.

• Microsoft Learn, "Class extension - Method wrapping and Chain of Command", updated 27 Mar 2026, accessed 18 Oct 2026.
• Microsoft Learn, "Customization Analysis Report (CAR)", updated 26 Feb 2026, accessed 18 Oct 2026.
• Microsoft Learn, "Design principles and best practices for data entities", updated 20 Jan 2026, accessed 18 Oct 2026.
• Microsoft Learn, "Write extensible code", updated 27 Mar 2026, accessed 18 Oct 2026.
• Microsoft Learn, "Testing and validations", updated 29 Apr 2026, accessed 18 Oct 2026.