In short, the biggest difference is the presence of humans in the loop. A system like Airflow has really nice ways to programmatically specify workflows in Python, but it assumes that all steps of the workflows are executed on machines. Orchestra, on the other hand, assumes machine step execution on workflow management tools like Airflow or Celery, but importantly also allows workflow steps to be executed by humans. As a result, our development focus is largely on the best user interfaces, review mechanisms, and collaborative hooks into tools like Slack/Google Drive for facilitating human & machine-completed tasks. I hope that helps!
-Adam