The usual answer, and its competitors
The UNIVAC I, delivered to the US Census Bureau in 1951, is the usual answer to the question of the first commercial computer, and it is a fair one if “commercial” means a machine designed to be manufactured and sold to customers rather than built once for a research project. It is not the only answer. In Britain, the Ferranti Mark 1 and the LEO were close rivals, and which one comes first depends on how you define the word. None of this happened in 1967, when business computing was already routine, and the article ends there for contrast.
UNIVAC I
J. Presper Eckert and John Mauchly, who had worked on ENIAC, left the University of Pennsylvania in the late 1940s to start their own firm, the Eckert-Mauchly Computer Corporation. Money was tight, and in 1950 the company was bought by Remington Rand. The machine they had been designing, the UNIVAC I (Universal Automatic Computer), was delivered to the Census Bureau in March 1951.
It was a large vacuum-tube machine that stored data on magnetic tape rather than only on punched cards, which made it well suited to handling large volumes of records. Around 46 were reportedly built over the following years, a small number by later standards, but enough to prove that a manufacturer could sell computers to a range of customers.
Two other points are often made about it. In 1954 an appliance factory run by General Electric in Louisville, Kentucky, became one of the first private businesses to use a UNIVAC, and it was used partly for payroll. And in November 1952 the machine drew public attention when a UNIVAC was used by the CBS television network on the night of the US presidential election. Its early projection pointed to a large win for Dwight Eisenhower, which the broadcasters were at first reluctant to believe, and the result turned out to be close to what the machine had suggested.
The British rivals
The Ferranti Mark 1 was developed from the experimental Manchester computers of the late 1940s, and the first one was delivered to the University of Manchester in February 1951, a few weeks before the UNIVAC. By a strict date count it can claim to be the first commercially manufactured general-purpose computer. Its customers were mainly universities and research bodies, so it was less a business machine than a scientific one.
The LEO, short for Lyons Electronic Office, has a different claim. J. Lyons and Company, a British catering and food firm best known for its tea shops, built its own computer, based on the Cambridge EDSAC design, to handle its internal accounting and stock calculations. It ran its first routine job in 1951 and took on payroll work in the early 1950s. LEO was not originally built for sale, though the firm later set up a company to market the design. As a pioneering business use of a computer, it is hard to beat.
So the answer depends on the question being asked. The earliest delivery of a manufactured machine may point to Ferranti, the first design built to sell as a product to many kinds of customers to UNIVAC, and the first computer built by a business for its own office routines to LEO.
What business used them for
Early commercial computers were expensive and difficult to run. They filled rooms, needed trained operators and were bought by organisations with large amounts of repetitive record-keeping: government statistics, payroll, insurance and accounting. The shift was less about doing new things than doing familiar clerical tasks faster and with fewer errors.
IBM, which had long dominated punched-card equipment, entered the market in the early 1950s. Its IBM 650, introduced in 1954, was widely installed and is often described as one of the first mass-produced computers, which helped shift the industry’s centre of gravity towards IBM.
The situation by 1967
By 1967 the idea of a business computer was no longer remarkable. The IBM System/360, announced in April 1964, had become the standard reference point for the industry. It was a range of compatible machines intended to serve both commercial and scientific users, so a company could start small and move to a larger model without rewriting its programs. Many large companies had a computer department, a sign of how far the American economy had come to rely on data processing, and IBM mainframes were found in banks, airlines, insurers and government offices.
Compared with the UNIVAC I of 1951, a machine of 1967 was far faster and smaller for the same work, thanks to transistors and early integrated circuits, and it was far easier to program through high-level languages like COBOL. The business revolution the title mentions was real, but it unfolded gradually over about fifteen years, rather than arriving at once.