We envision it in a similar way. It will be a separate tab, "Data catalog" with a list of data entities. When a data entity is selected, the user provides parameters and the output format - xls, dset, or txt. The result will remain available for some time (e.g. 1 hour, configurable in space settings). Alternatively, a download link to the result will be sent when the task finishes.
Although, in the 1st rollout phase, the Server UI won't be ready yet. The data catalog will only be available via Desktops in the beginning. The Server UI will be added in later phases.
A sample data preview will be available for each data entity (as metadata) so that users can get an idea about what kind of data they would get from this data entity. A preview of result datasets is not planned.
I believe, the bi-directional export feature should also be interesting to you. It's just very unusual for a data catalog, so it may not be immediately clear what use cases it may have. But it's actually quite cool, as it enables non-technical users to securely and reliably export/update data in enterprise systems without knowing SQL or web APIs. You will also no longer need to build web-forms or applications to let users see/modify data in enterprise systems.