I'm a technology guy, but one lesson I've learned is that not all solutions can be solved by throwing technology at it. For example, try classifying a first name as male or female. Quite a few are pretty clear, but there are enough that aren't to make it something that you just can't hit 100% accuracy. Sometimes you need the proverbial fuzzy search, sometimes you need a person to look at it, sometimes you can use technology to make it easier to do that, sometimes you just don't need much technology.
Fuzzy writing? My point is that you have to find solutions with enough accuracy, not perfect accuracy, and with an appropriate cost. It's what I phrase as good, better, and best. Sometimes you need best, a lot more often you just need good!
Now part two of this post is that we all hear a lot about service oriented architecture, web services, services in the cloud, etc. They seem cool, but we don't seem to see a lot of them getting used in the corporate world beyond trivial examples like retrieving the weather for a given zip code. I'll grant that mashups around Google Maps and similar have gotten some traction. Maybe I just don't read the right blogs!
So for a real world example that combines those two points, take a look at this case study that uses the Amazon Mechanical Turk. My friend Jon Winer was one of the key players in the implementation, and it was interesting to hear over many months his evolving understanding of how to use it, and then how to use it effectively within his company (he didn't share any secrets). It's a digitally implemented analog solution if that makes sense, that uses a web service as a key part of their business to answer questions like "is this image a printer?". Side note, my friend and blogger Kendal Van Dyke works at the same company.