Room 3 · The Winters · 1980

XCON, the Expert System That Made Money

XCON, the Expert System That Made Money (John McDermott · Carnegie Mellon and DEC, 1980): AI comes to the factory.

XCON (also called R1) configured orders for Digital Equipment Corporation's VAX computers, checking that each order had compatible components. It went into use in 1980 and was credited with saving tens of millions of dollars a year.

In the late 1970s Digital Equipment Corporation, DEC, sold VAX computers that customers ordered in a huge number of combinations: processors, memory, disks, cabinets, cables. Every order had to be checked by technical staff to make sure the parts fit together and that nothing was missing. Errors were costly and common.

In 1978 John McDermott, at Carnegie Mellon University, began building a system to do that work. He called it R1. Inside DEC it was later renamed XCON, for eXpert CONfigurer. It was written in OPS5, a language for rule-based programs, and it captured the knowledge of DEC's configuration experts as rules of the form "if the order is in this situation, then do this next." McDermott described it in 1982 in the journal Artificial Intelligence, in a paper titled "R1: A Rule-Based Configurer of Computer Systems."

XCON entered regular use at DEC in 1980. It grew steadily, reaching thousands of rules as new products arrived, and DEC reported that it saved the company tens of millions of dollars a year in errors avoided and work saved. For a field just coming out of its first winter, that was decisive evidence: AI could pay for itself.

XCON's success, together with systems like MYCIN, set off a wave of corporate investment in expert systems in the 1980s. Companies created AI departments, startups sold expert system tools, and specialized hardware appeared to run them.

It also revealed a weakness. Keeping thousands of hand-written rules consistent required a permanent team, and every change could have side effects. That cost of maintenance is part of the story of the second winter, later in this room.

Ask your AI: what kind of problem today would still be better solved with explicit rules than with a learned model?

Did you know? Its success pushed many companies to invest in expert systems during the 1980s.

Source: doi.org/10.1016/0004-3702(82)90021-2

See it in the museum →

← The Lighthill Report · The Fifth Generation Project →

All of Room 3: The Winters

Cite this: Nazareno, I. (2026). XCON, the Expert System That Made Money (1980). New Mind Museum. Retrieved November 17, 2026, from https://newmindmuseum.com/museum/xcon-the-expert-system-that-made-money-1980/