Hi David,

I would think of this in the same way you would were you to do an insert using the DB tool of your choice outside of Pega. (each of the SQL statements are atomic)

1) insert

2) key is generated in row

3) how do you get it back using SQL?

The problem is that the row is uniquely identified by the key, the whole key, and nothing but the key (lol). So you would need an approximation built on the data you have (without knowing the key) , and then get the max createdatetime or some other value outside of the key that would allow you to sort the result set. This is prone to errors though since another transaction may have inserted between the time you committed the insert and the subsequent select - (substitute Obj-Save for the insert and Obj-Browse for the select)
 

Alternately you could use a stored procedure to accept the data as input and return the value of the unique key (this assumes the database has the ability to return to you the just generated unique key within the scope of your transaction)

Sorry for the long winded response - take these with a grain of salt and consider them when deciding what to do. There may be other solutions that are better.