If the source dataset is not very large you can unpivot, apply a rule, then pivot back the entire dataset at once. See the example below:
dynamic-type-conversion.morph (5.5 KB)
Book1.xlsx (8.9 KB)
In the example I deliberately included some values that fail data type conversion. You can later apply a filter or the “Halt on condition” action to do a data quality check. I’ve also added a dynamic renaming step (not sure if it was required or not).
If the dataset is large, then you will have to split it in chunks (e.g. by 10000 rows) and then iterate.