Quickly why they don't work:
You create a huge chain correlational assumptions. First that visual-spatial tasks of this kind predict performance on visual tasks. 2. That performance on visual tasks predict general intelligence (whatever that is). 3. That this notion of general intelligence (which is usually and arbitrarily defined not to include social skills) actually correlates with the tasks that you think the person will be performing, and finally that your idea of what the role has an impact on the company. Of course it is completely absurd, what they are selling is snake oil, plain and simple.
The remedy I recommend is simple, talk to the person - do it and you will be able to tell within 5 minutes.
There is a strong correlation between IQ and professional achievement whether you want to believe it or not.
First, if you're just hiring assistants from the Philippines they could just have someone else take the test or get around it some other way.
Second, you have no good data to support this hiring practice. You're free to use it, but it's no better than just hiring a random person from your pool of applicants. You might as well screen based on their favorite color too to just make up filtering criteria.