We can't propose effective solutions to the conference imbalance until we accurately identify the source. Your post assumes that the problem is with the function being used to evaluate speakers. That's far from certain. An arguably much likelier explanation is that, due to the huge gender imbalance in computing, the percentage of female applicants to speak at a conference is much lower than 20%. Therefore, the percentage of qualified applicants who are female is presumably much lower than 20%.
Let's assume for the sake of this line of reasoning that the function for evaluating a speaker's ability is accurate. If you are looking to invite ten speakers to your conference and there is only one woman among the top ten speakers, it would make perfect sense if you only invited one woman. That makes you good at your job, not sexist.
Now how do we go about changing this? Well, one approach is what you're suggesting--affirmative action. There are certainly pros and cons to introducing such a minimum quota, but I find the article's assessment that the cons seem to substantially outweigh the pros to be very persuasive. Inviting inferior speakers in an effort to increase the representation of a specific group at a conference is problematic for two main reasons: (1) it results in a conference that is expected to be worse quality and (2) it deprives women who actually are evaluated to be in the top ten speakers from knowing, and from having other people know, that their value to the conference came from their speaking ability rather than their second X chromosome. Additional ways in which the quota approach is problematic are that (3) it discriminates against the many other groups that are under-represented at tech conferences, as well as the oft-debated argument that (4) inviting a worse female speaker over a better male speaker is unfair for the man.
If we suppose the quota system does more harm than good, what alternatives do we have? That's the million dollar question! The only answer I've got is to focus our efforts on addressing the sources of women's under-representation in computing in general, such as encouraging female students to study computer science, getting schools to examine whether their classroom environment discourages female students from pursuing CS, and help software companies identify whether their workplace atmosphere discourages women and what steps they can take to address that. Forcing conferences to invite more women or businesses to hire more women is a highly problematic and very short-term hack. Identifying and fixing the structural barriers to women in computing is a highly difficult but very long-term solution.