It is human who decides of the opening, of the strategic plan and some critical hits. The machine will be much more powerful in all aspects of the game tactics, trade with gains of material and find the best way to go to mate.
The best companions of analysis that I found are : Houdini 2.0b, Houdini 1.5a, Rybka 4.1, Fritz 13, Critter 1.2 and Strelka 5.1. These choices of chess programs have not done randomly, they each offer a different special. For example : Houdini 2.0b with his predecessor Houdini 1.5a will complete the selection of candidate moves. Rybka will allow me to save time by his ability to find in the middle of some good moves in less than 5 minutes. Fritz 13 is very strong in the final of Towers and in some aspects of positional play. Strelka offers me both moves that seem crazy at first, but are powerful in certain positions. Critter is the last machine I use, it is there to comfort me when more equal opportunities.
Then there is the choice of opening books. Before I made my own opening trees, now it has become impossible because of the strong programs that have transformed to the same coherent structure of a successful implementation to arrive in mid-game. So my choice is shifting to specialized and general books. For example : The book of shredder which is complemented by its function online, and various books like atlantis 7.8 and arilovsu 9.8. This full advantage with the databases that I currently have parts of FIDE and ICCF.
And it's the same as for engines analyze, do not make choices at random openings. A good study of the structures of pieces/pawns must be made in advance, because in the final, it does not forgive and defeat can come quickly. However powerful chess software, you can lose because of stupid structures.
The share of human choice is important. Correspondence chess has nothing to do with the matches between programs. And if sometimes I ask questions here that may seem silly to those familiar with this forum. My research is that performance goes beyond the "Houdini beats all other programs in 1 minute". And programs react differently in the long time control. Took me a while to trust Houdini 2 and that is why I always complete analysis with Houdini 1.5a. If the analysis of Houdini 2 is different from that of Houdini 1.5a, then I use Fritz 13 and Rybka 4.1. And if I find Houdini 2 was wrong, then I plug in his office "learning" to teach him a better move.
Without human intervention, you will struggle to exceed the 2200 elo in ICCF. But without the machine you will never reach 2000. That is why at the beginning of my article, I say that the human and machine are one now in correspondence chess.
I hope this article will help you in your future choices. The computer did not kill correspondence chess, but made it different. And points beyond 2200 are fighting very hard to win.