Approaches like Zero ETL e.g. at BigLake and also at AWS Redshift also allow direct queries to source systems. Whether this is always sufficient or whether extract and transform must take place certainly depends on the use case — at least in my opinion….