I get asked about the future viability of transcription companies a lot, "Don't you think that transcription will be done by computers or outsourced to cheaper labor markets in the next few years?"
Absolutely I think that certain transcription work will be relegated to that cheaper model. It's already happening. There are many transcription services that rely on huge volumes of work and razor-thin profit margins, charging as low as $40-60 per hour of audio. I'm not criticizing them. They occupy a necessary niche in the market, just as transcription firms like Rogers Word Service occupy an equally necessary niche. I'm not sure how they get their transcriptions done, but I do know that very few of our transcribers would be apt to accept the sort of pay that $40-60 per audio hour allows for (likely less than $8-9 an hour.)
The thing about transcription is you don't quite realize how difficult it is until you do it. Transcription is skilled labor. Not everyone can do it. It takes a special kind of attention to detail and ability to concentrate and listen that is actually quite rare. To be honest, more often than not, the transcribers who take our test do not pass. And these are almost universally folks with years of transcription or writing experience.
When I'm transcribing, there is a constant stream of "small" decisions I have to make. I have to decide whether to use a period or a semicolon, depending on how related or dependent the two clauses are. Because of greater context clues as the discussion goes on, I might need to go back and fix an earlier part of the transcript where I wasn't quite sure what they were talking about.
There are dialect and regionalism questions to wrestle with. The first time I heard a British speaker use the term "bog standard," I had to spend about 15 minutes on Google, trying to figure out what the heck she was saying. Once I had that information, everything made perfect sense. It's just not a phrase people in North Carolina say, as far as I can tell.
People don't speak in complete sentences or with proper grammar. How do you render the spoken language into clear, readable written form while not changing the actual words? It's surprisingly not always clear where one sentence ends and another begins. There are situations where a speaker's thoughts are constantly connected by conjunctions (and, but, or). Where do you decide, "Okay, this is enough. I'm just going to have to start a new sentence with a conjunction because this is getting unreadable"?
Which "you know" or "like" is just verbal tic, and which one actually conveys some meaning? It's a judgment call, and it's usually up to the editor or the transcriptionist.
These are just a few examples of the things that go into making a transcript the very best it can be. And I think you need a decently paid, experienced transcriptionist to make those decisions and to care enough about the final product to make it happen.