Well, we are at a stage now that we need to deal with considering a huge amount of highyl specific, mainly scientific, papers into various languages and from various languages into English. It is not us who are the customers - we should just think about the what, when, who, how long, price etc. At this stage we need a huge network of translators that tell us in a detailed way which exact skills they have. In the case of the potential customer outsourcing to an agency is not a good idea, but working with a PM who organised the translators for that company.
Reasons: the papers are highly specific of any kind of science. Often a direct contact between author and translator will be needed. The contents is free content, so budget is not the highest even if also not the lowest. Agencies will not be able to deal with such a way of working, because it implies much manual work by the PM which means relatively high costs. We need to know each single person. Huge agencies often outsource to smaller agencies that handle that specific language combination (sometimes it is about rare languages ... not easy to find translators) or a specific area (for example medical translation). The smaller agencies often cannot say "sorry, we cannot help you", because otherwise they could loose an important customer. They need to find someone who does the job. In the worst case the second agency does outsource to a third agency in a low-price country (I don't say all do it - don't get it wrong, but some do it ... and you don't know who are that "some"). Then you have a translator working on a text where he very likely has questions - these questions then must go back the whole chain to find answers and then these answers must follow its way back to the translator. For certain kinds of translations: this is simply not the way it may go. Another thing that is to be considered: translators must live on their work. If the texts are highly specific they need time for research - so how much can they do in an hour? How much can they earn?
We are translators, not scientists, nor engineers - finding a scientist who is a translator or an engineer who is one: that is a reall rare case ... these are the jewels for very specific texts and finding them means knowing many people, passing word of mouth, asking, communicating, reading some of their texts.
We need people who do not worry to say "no thank you, this is not a text for me" if they cannot deal with a job - only working in such a way these people will be able to be part of that community. And: we will need a huge community of translators to cover all the needs. It will take time to create all these connections, but on the other hand: in that period needed we will create one of the most particular networks of translators that are around - different to anything that you know by now.