I should probably clarify that this is usually done only for a gene or two at a time and is quite labor intensive.
First you use CRISPR to target the original gene, causing a cut which then gets improperly repaired. You screen a bunch of edited cell populations for something where the repair causes an insertion or deletion of nucleotides in the gene that disrupts the translation to amino acids (as amino acids are encoded by patterns of 3 nucleotides at a time, adding or subtracting a number non-divisible by 3 will change the identity of all of the amino acids downstream of the edit). If there is a protein produced after this edit, it will be scrambled and broken down quickly.
To put a gene back into a cell line you use lentiviral transduction to incorporate it into the genome. Lentiviruses are viruses like HIV that use a reverse transcriptase to edit your genome directly. The ones used as genetic tools lack the ability to replicate, and contain a payload of DNA that you want to encode into the cell lines. Usually the plasmids for lentiviruses encode a resistence gene for a cytotoxic drug such as puromycin (inhibits eukaryotic protein translation), so that when you grow the cells in puro, only those that are expressing the lentiviral DNA will survive (also this keeps the cell line from eventually kicking out the viral DNA).
Expressing an orthologous gene from other species is a common technique for validating function. If you can demonstrate an effect from knocking out a gene, and then reverse the effect by adding back in the gene from a different species (usually will be slightly different), you have grounds to make a claim on the function of the gene product.