Source: p 287, Sweet Reason: A Field Guide to Modern Logic (2010 2 ed) by Henle, Garfield, Tymoczko.
Predicate is the reason we started on deductions. In Sentential, remember, we can verify that an argument is valid by using truth tables or the short-cut method. With Predicate, we have no such tool. We can show that an argument is invalid if we exhibit a universe in which the premises are true and the conclusion is false, but (until now) we have had no means of showing that an argument in Predicate is valid.
In the 2 Logic textbooks read (the above and Hurley's), each exercise on deductions (e.g. the following) divulges the (in)validity of the argument, and so enables you to know instantly what to do. But without computers, for arguments with > 10 long, convoluted premises: what if the (in)validity were concealed or tacit?
Per the above, I already know that a Valid Argument can be proven only with Deduction, and an Invalid Argument disproven with a Counterexample. But before 1. writing a Deduction or 2. finding a counterexample, you must have conjectured the argument, respectively, 1. valid and 2. invalid (Otherwise, how would you know to write a Deduction or find a counterexample?!?) So how would you conjecture (in)validity before doing the work?