Test a workflow with a small known fixture

A fixture with known answers makes it easier to detect mistakes before using a workflow on important material. Include an ordinary case and one meaningful exception.

Start with a small handoff

Select a role by its working guide and deliverables, then verify the connected AI tool and workspace. Keep the first run small enough that you can inspect every output. The desktop needs to stay running while work executes. Marketplace sign-in, model access and file-generation tools are separate setup requirements; a success in one does not prove the others are ready.

Inputs for this workflow: A small fictional or sanitized dataset with manually known results.

Work through the task

1. Write the expected output first.
2. Run the intended brief on the fixture.

A brief you can adapt

Summarize this fixture, count accepted and rejected records separately and explain why each rejected row was excluded. Do not infer missing values.

The filenames above are examples. Replace them with your actual inputs and destination before submitting the task.

Review the deliverable

Compare every output value with the expected answer and inspect the exception handling.

Where this approach can fail

Clearly label fixture data as illustrative. Do not publish its numbers as real business performance.

For the related desktop setup, see first task.