Use AI tasks and understand extension boundaries
26R1 reference · Documentation-based tutorial draft · Not executed in IFS
IFS AI Task is a specific integration
Section titled “IFS AI Task is a specific integration”An IFS AI Task invokes an available IFS Industrial AI use case and model. It is not an arbitrary prompt box connected to any external AI service.
Confirm that the business use case, model, required data, and IFS tokens are available in the target environment. Do not assume examples for another installed component will be selectable.
Configure a controlled AI step
Section titled “Configure a controlled AI step”- Choose the business use case and model.
- Inspect the model’s required parameters and map them explicitly.
- Give raw output, selected outputs, status, and error variables distinct names.
- Use JSONPath mappings matching the actual response.
- Add a decision for failed or unusable output.
- Test with fictional data before allowing the result to update business records.
If Log Errors is enabled, execution can continue after an exception; check the error/status before using the output.
Troubleshooting, normal invocation, and retries consume IFS tokens. Count expected calls and retry behavior when assessing a design. More retries are not a substitute for correcting invalid parameters.
Use AI Invocation History for operational evidence. Model output should be validated against the business requirement before it becomes an authoritative field value.
Workflow engine access is still governed by IFS
Section titled “Workflow engine access is still governed by IFS”The extension documentation describes access to underlying Camunda REST interfaces through IFS security. It does not mean every generic Camunda tutorial, endpoint, or BPMN feature is available unchanged.
Discover the environment’s supported endpoint and authentication route. Workflow External/Internal grants and administrator-only operations remain relevant. Avoid shipping a hardcoded engine base URL.
When custom Java is involved
Section titled “When custom Java is involved”Custom delegates and listeners are code development. The reference specifies IFS transactional base classes for delegates/listeners, plus compatible build and deployment requirements.
A class existing at deployment time does not prove its business logic works. Validate its transaction behavior, exceptions, required libraries, and target runtime through the solution’s development pipeline.
| Artifact | Responsibility |
|---|---|
| BPMN | Process model and task configuration |
| Delegate/listener Java | Custom task or lifecycle logic |
| Required libraries | Build/runtime dependencies |
| Trigger configuration | How the process starts |
| Grants and deployment records | Who may execute which version |
Keep these dependencies together in the change plan. Most first workflows should use supported IFS tasks before introducing custom runtime code.
The older extension page includes an authentication note that conflicts with the dedicated REST Task page. For current REST-task configuration, follow the dedicated task guidance describing Bearer and OAuth 2.0 client credentials, and verify your installed release.