General question on batches and Azimuth

[gsudre]

Hi, after reading the traditional Seurat integration vignette and the Azimuth FAQ on different batches, I'm struggling to understand what the workflow should be. Am I wrong that Azimuth does not take into consideration the previous clustering results? For example, say I run the integration tutorial, and none of my clusters seems to be related to technical or biological artifacts. Then, I use Azimuth to classify my cells. Couldn't I just have skipped all post-QC steps and gone straight to running Azimuth for each batch individually, and then just continue the analysis from there (e.g. pseudobulk on Azimuth results for cluster A of each batch, then cluster B, etc)? I think I might be missing something! Thanks :)