AI assistance and integrations

The academy provides readable explanations for people and AI tools. It does not grant access to a company's records or authorize operations in OSERP.

Ask a useful question

Include the screen name, business task, current status, and relevant non-sensitive context. For example: “In Item Master, what is the difference between Standard Price and MRP?” or “My journal date falls outside the configured periods; which setting should the administrator inspect?”

Ask the assistant to cite the relevant academy page and state uncertainty when a behavior is not verified. A generated answer should not invent a field limit or claim a workflow is implemented because a configuration flag exists.

Company and item knowledge

Item Master includes Markdown content and document-indexing choices for AI use. Decide what information is appropriate before indexing it. Product instructions and specifications are different from employee data, customer financial records, credentials, or private implementation notes.

The /ai screen is associated with settings management in the menu table. Provider/model configuration and usage settings belong to authorized administration. Do not paste secret keys into the public academy or into a support example.

Read versus act

Reading a public procedure is different from executing it. An AI or integration performing an operation must use an authenticated, authorized application interface and the correct company context. Documentation retrieval does not bypass role checks, posting validation, approval rules, or period locks.

For consequential actions, inspect the target record, proposed change, and expected outcome. After execution, verify the actual saved result and retain its reference. Never treat an assistant's statement “done” as a replacement for application evidence.

Integrations

OSERP has backend routes and MCP-related implementation, but this academy edition does not advertise every internal route as a supported public API contract. Use the published integration interface approved for your deployment. Authentication, scopes, versioning, retry behavior, and supported operations must be documented for that interface.

If a request times out, inspect the operation's supported idempotency/retry behavior and check for an existing record. Do not assume repeating an arbitrary POST is safe.

Continue to AI-readable academy content for the public knowledge exports.