An organization gets the software systems it deserves

You can't sprinkle magic computer dust to bring order to chaos

Around 1990, I was at Freddie Mac, the mortgage giant. It had gone through a period of very rapid growth, and its computer systems were in an awkward position.

What was true at Freddie, and at many other large organizations at that time, was that personal computers were starting to have an impact. Information systems divisions did not have a model for managing the use of them. A typical large enterprise had some key business systems that ran on large, mainframe computers. But various departments had adopted networked personal computers for document processing and spreadsheet calculations. Inevitably, some of these “end users” found it easier to set up small databases and applications that ran on personal computers and never were connected to main systems.

At Freddie, this was a nightmare for the accounting department and for the information systems division. The accountants did not have direct access to the end-user financial systems. The information systems division had no control over the software being used.

As I watched the company struggle with these issues, I noticed something. Some divisions, notably the one that handled payment processing, were very strict and precise in the way they implemented policies and procedures. Other divisions, notably the one that negotiated agreements with mortgage lenders, were very creative and ad hoc in the way they adapted contracts to customer needs.

Payment processing had computer systems that were reliable, well documented, and tightly integrated. Contract negotiations had information that was unreliable, undocumented, and scattered.

I could not help but notice the correlation: tight organizations had well-engineered computer systems; loose organizations had messy data. It struck me that

Organizations get the computer systems they deserve.

When you have an organization that is improvisational and fragmented, its computer support will not be well-engineered. Moreover, top management cannot simply order up a well-engineered computer system for a loose organization. You cannot use information technology to organize the disorganized. The software engineers cannot drive the business from the back seat. That was a difficult lesson for Freddie Mac’s senior management to absorb.

Instead, if you desire a well-engineered computer system, you have to tighten up the business process. Managers have to be able to articulate business rules, write them down, and stick to them.

Electronic Medical Records

About 15 years ago, policy makers began to champion electronic medical records as a panacea for controlling health care costs in the United States. But that does not take into account the fragmented nature of the underlying process.

Doctors have been trained differently. They experienced different populations of patients. They have worked in disparate settings, with different types of support. They have different styles: some are very methodical, while others are more intuitive.

If health care were standardized, then electronic medical records could bring large benefits at relatively low cost. As it is, the costs are high and the benefits do not seem to be realized.

We may see some standardization as the health care industry consolidates, with small practices becoming extinct. Part of this consolidation is due, in fact, to the high cost of complying with regulations requiring electronic medical records. Small group practices cannot afford the software and the training involved.

The Bottom Line

Computer technology is not some magic fairy dust that you can sprinkle on a loose, variable business process and turn it into a consistent, well-oiled machine. If a process is highly standardized to begin with, or managers undertake the effort to exert more precise control, then it can be usefully represented using data. Otherwise, an attempt to impose the order of a computer system on the chaotic organization is unlikely to succeed.

In this context, it is interesting to watch the large language model approach to artificial intelligence. Humans’ use of language is far from completely standardized. When chatbots make mistakes and behave inconsistently, they seem more human and more “creative.” But will they succumb to the imperative of computer systems and be designed to be predictable?

This essay is part of a series on human interdependence.