An information system is to an organization what a service infrastructure is to a building. The surrounding entities, the building or the organization, would likely not function without their respective systems. Unlike the building or the organization, we don't acknowledge one of these systems as its own entity, but more as a collection of interacting parts. Furthermore, it is not the features, technology or cost of these systems that determines their success, but that they exhibit the appropriate behaviour. When we set out to acquire or improve these systems, while we can make the job easier by using prefabricated components, we must often configure them in a way that is unique to the building or organization in order to achieve the behaviour we're looking for.

When we can reduce a job or piece of functionality to a well-established shape or series of steps, it is customary to turn it into a component, machine or tool, provided it is economical to do so. The cost of making this tool is offset by the permanent benefit it provides. If somebody has already made an appropriate tool, it is often cheaper to buy one rather than make our own. However, if a tool does too much, it tends not to do much of anything well, and using the wrong tool for the job may have unwelcome results. Imagine trying to convince a tradesman to use a power saw as a hammer, or swap his hand-picked kit for the Swiss-Army equivalent.

In an information system, a corkboard in a company lunch room may be just as important to the functioning of the organization as a Web site or database. The technology used is irrelevant, the only important thing is that it works. You might entertain replacing an item like a corkboard with some newer technology, though, if it was failing to perform in certain ways, such as people missing messages or posting them back to the corkboard faster than you could handle them. If you want to improve volume, access and automation, it may be time to consider computers.

In all 21st-century businesses there is a kind of work to be done that deals exclusively with information, from delivery notifications to strategic planning. Some parts of this work are trivial, others business-critical. Some are time-sensitive, others demand accuracy. Again, if we can describe a part of the process accurately enough, we can turn it into a machine. If somebody hasn't done it already, we can make our own, provided it is economical to do so. For working with information, there is no more economical a medium than software, as it is literally thought turned into action. The variable costs to make it are zero and the capital costs are close to it. The only other cost of making software is the cost of having somebody type it out.

Therein lies the catch: Software is only cheap to make if you know what to type. If you don't, you'll just have to keep doing it over until you get it right, and somebody has to pay for that. That, or product gets left unfinished, potentially causing more problems than it solves. What gets written into software of this kind is ideally some part of the method and capability of a business, but it is always mediated by the opinion and bias of the person writing it. If this person doesn't understand what the business needs to accomplish, he'll never get it right. You can't fault him for that either, really, because he has more pressing things to worry about, like making things work at a technical level. What you can fault him for is trying to sell you a solution that solves all of his technical problems and none of your business problems.

If you can figure out what the business is trying to accomplish from the perspective of its goals and process, before considering technology, the person writing the software can do a lot less guessing and filling in of blanks with his own assumptions. This means he'll get things closer to right, closer to the first time. That preliminary work, called design research among many other names, concentrates and expresses the information necessary to describe the business process in a way the developer can understand, and provides a benchmark against which we can compare his results. Since business processes are often much longer-lived than technological ones, so are the results of this research, which can be expanded over time and applied to other parts of an information system. The value of this research is that of self-awareness on the part of the organization, and the establishment of a home for an essential aspect of business intelligence.