"The problem with being a helpful, efficient but largely anonymous middleman is pretty obvious. Someone can come along who is cheaper, faster and more efficient. And that someone might be the customer aided by a computer."
Straight-through processing - sometimes seen as one of the potential sources of value from service-oriented architecture - is a mixed blessing, as I've pointed out before. There is always a danger of disintermediation, especially if the added value provided by the middleman appears trivial or easy to replicate. As Seth Godin points out, if you are just providing an anonymous human approximation of Google, don't bother - Google does it better.
In some industries, the middleman had become lazy - taking a cut for not doing very much. My local travel agent never gave me any useful information or advice, merely handed out brochures from the travel companies and struggled with the complexities of the booking system; so I stopped using them. And in the financial markets there are still companies that think they deserve a percentage of your pension fund in return for pressing a few buttons now and again.
In the past companies like these have often been able to take advantage of a strategic position in some larger process, extracting "rent" but without creating value. And there are doubtless still many niches in the system ecosystem that can be exploited in this way.
However, service ecosystem niches that allow companies to draw rent without creating value are going to be short-lived, and rightly so. In hard times, the percentage should only go to companies that are providing genuine value. (Actually the same principle applies all the time, but people are more motivated to follow this principle now than they have ever been before.) And ecosystem SOA should help us to make sure of this.
Then business strategy in a dynamic service ecosystem should not be based on finding and maintaining positional advantage, but on creating and maintaining genuinely productive relationships.