I pulled up the book to get the specific details. ("IBM's Early Computers by Charles Bashe, (c) 1986 by MIT Press.) From the book:
The government and industry needed a fast way to transmit data from field offices to headquarters.
In 1941 IBM introduced the 40 Card punch which read telegraph tape and converted the characters into punched cards. The 57 tape punch read cards and punched telegraph tape. Telegraph interface was used instead of phones because of the urgency of the government's request for the gear and the widespread availability of telegraph facilities using paper tape facilities.
In 1954 IBM announced the 66 Data Transceiver which used 4000 Hz voice grade phone lines and allowed transmission of eleven cards per minute, up to 16 characters per second. IBM developed its own self checking 8 bit code. The unit had what was later called a modem. The 66 was initially restricted to leased and private lines.
In those days, the vast majority of data was processed on card tabulating machines, not electronic computers. "Real Time" meant a clerk looked up an account in a tub file.
In 1960 IBM announced the 7701 Magnetic Tape Transmission unit which utilized the full 1200 bps data rate of voice grade lines. This used synchronous transmission which meant individual characters didn't need their own start/stop control characters.
So, to answer Pat's question, IBM had a working modem on the market by
1954. How many were sold I don't know. I don't know what AT&T was doing, though obviously they were aware of this need. AT&T was developing dialble private telephone networks for very large organizations in those days. Perhaps if someone has Bell Labs Record magazine for the 1950s there would be more information.I suspect that in 1954 the cost of a voice grade line, whether dial up or private, was much higher than telegraph lines, so the demand for modems for voice grade usage was probably low. Large organizations had couriers going between remote plants; it may have been cheaper to simply ship documents to HQ for processing rather than key them in remotely. In those days businesses had to make do with delayed information; the cost of speeding it up was just too high. Things moved a bit slower in those days; purchase orders took a few days to be confirmed and there was a longer lead time between order and delivery due to processing. I believe private teletype networks were widely employed in large organizations and orders and status were transmitted via that mode. As Pat described in the credit card bureau, large companies had armies of clerks to prepare and then process this information. In other words, I don't think machine-to-machine data transmission was used much in the 1950s.
I sit here at my PC and realize how far we've come. At the click of my mouse in literally seconds I can be on a business website and order virtually anything I want and it will be shipped out tonight. If this were the 1950s I'd have to use a catalog, printed once a year and obviously not amendable to changes during the year. I'd likely write my order and mail it, hoping I calculated shipping, etc., correctly; if there was a mistake I wouldn't learn about it for a week or so. The catalog center had its armies of clerks and pickers. Sears had a huge building in Philadelphia as a regional center and employed many people. The people was blown up to become a strip shopping center. I feel bad for the loss of so many jobs, but I realize I'd rather use faster methods.