For explaining my doubt, I am taking an example of a table named "Employee" in the database named "Company Employee" table has an autonumberd primary key column "EmployeeID"
Auto-numbered field cannot be updated from the client-side. Eventhough this number is generated when creating new DATAROW in the DATATABLE, this value has to be ignored when doing an INSERT (thus, the whole idea of auto-numbered field in the database - database has to assign that number).

Even if you don't set the EmployeeID as auto-numbered, you probably will set it as PRIMARY KEY. Therefore, when the last user inserts new record w/ EmployeeID = 100, it will throw an exception. So, it is up to you as the programmer to handle this exception gracefully.

Can anybody explain me what exactly hapens in database whn both of the statements executed?..
Well, the first user that executes the update/insert command will save all the changes in the database. The last user on the other hand will probably get some kind of error. Again, it is up to you as a programmer to make sure to handle data integration. For example, when doing an update, check how many records are you updating and then check back how many were actually updated.

Why is this better? Well, actually it is only better to the user but unfortunately really sucks for the programmer. But then, we program for the user, are we not?

So, with the disconnected model, data integration must be handled by the programmer(s) instead of the database (most of them).