Doing It Backwards

I believe Paul Graham's astute analysis, of why Americans make the best software and why the Japanese make the best cars, can help point to a solution to the greatest technical problem faced by insurance companies today.

Most medium to large Property and Casualty insurance companies desperately want to modernize and improve their computer systems. Most, if not all, of these companies have been frustrated in their attempts to do this. I believe the problem may be because they are trying to do it backwards.

For shorthand, I will refer to Paul's two approaches as Careful-Design and Just-Do-It. I agree with Paul's conclusion that the best software is created using the Just-Do-It method. This identifies part (but not all) of the reason insurance companies have been frustrated. Most of the projects upon which insurance companies embark in their quest to improve thier computer systems tend to use the Careful-Design method. A lot of careful, systematic planning is put into gathering requirements, spelling out specifications, etc. Many of these projects either end up failing, or at the very least, taking much longer than anticipated.

But, that's not the whole story. An insurance company's computer system is not one huge piece of software. It is a large collection of programs, most of which need to communicate with each other and with the outside world. This need to communicate, or integrate, is one of the biggest problems such systems need to deal with. How do most software projects deal with this integration problem? With the Just-Do-It method, of course!

Even when a company purchases an excellent and extremely useful piece of software, they usually spend enormous resources in the attempt to integrate it with their other systems. Usually, this is done with the Just-Do-It method. Almost invariably, this takes much longer and costs much more than it should.

I believe the integration of multiple systems, both within a company and with the outside world, is a problem that needs the Careful-Design approach. The Just-Do-It method is just not adequate here because it is not really software creation; it is infrastructure-building, and has a lot in common with other infrastructure endeavors like roads, bridges, sewers, telephone lines, etc.

The best approach I have seen so far on how to build such an integration infrastructure is the work being done by Sean McGrath with the Irish Government. His company has published a couple of excellent White Papers on the subject.

Unfortunately, most companies just don't take the time to think about investing in infrastructure. It isn't really the money. They just don't seem to understand the need. (It would require a sizeable paradigm shift in their thinking.) They want a solution to the specific problem that they are dealing with today. Whether they buy or build a new software system, when it comes to integrating it with their other systems, they want to Just-Do-It.