We think (and so define in IQ calibration) that mean intelligence is the same for men and women.
And it's that calibration that's problematic here. The brains of men and women are typically very different, making them excel at different types of tasks. The modern IQ calibration manipulates weights for these tasks to give both an average score of 100. The result is politically correct, but makes IQ an even more worthless measure than it was before gender-balance calibration was introduced.
The other problem lies in people assuming that the average for a gender tells you anything about a particular individual. If women are better/worse at task X, this doesn't mean a woman who applied for a position that requires X is better or worse than a man whom you can pick or not over that woman.
The result? Giving preferential treatment to either group is wrong, and will hurt not only the group you discriminate against, but your profits as well. No matter whether your task is a biology researcher, a lumberjack or a kindergarten teacher, the only valid method of choosing is being totally gender- (and race-, and so on)-blind. That woman who applied for that lumberjack job? She probably has a clue what she does, and thus deserves a try at the chainsaw. This kind of self-selection is not free of biases, but it makes comparing averages for men-vs-women (or blacks-vs-whites-vs-polka-dotted) pointless.
Yes, such selection of merits will make your team not represent the diversity ratios of the general population -- that's expected.