FARNBOROUGH AIR SHOW: On a picture perfect English summer day, the day was not filled with military airplanes, aside from an A400M. But nowadays some of the most important things in aerospace are invisible and intangible, like software and artificial intelligence. In fact, the most interesting thing the US defense press heard today about was how the Air Force is considering changing how it buys software to better make use of AI and machine learning.

“I was raised to see software as a product, but it looks much more as if software is a service,” Will Roper, head of Air Force acquisition told reporters. Roper has some ideas about how to buy software in weeks instead of months — but our discussion was all off the record, so you’ll have to wait to hear what he decides.

In addition to moving too slowly to buy software and develop code, Roper worried that the service is not “attracting enough people” who can write the code and algorithms to make use of AI and machine learning. “I contend that the companies driving AI are a different breed of company than those who drive evolution in hardware, especially companies that drove the hardware that have gotten us to today’s military.” And of course the pace is radically different: Today’s software “is done in month cycles, not year cycles.”

Ensuring the service gets software that really gets used is a key goal of Roper’s. That means putting airmen and coders in the same room, so the tech geeks know what the warfighters actually want and the techies can explain what’s actually possible. Will this result in more operations like the Kessel Run Experimentation Lab in Boston, where coders join airmen to build useful software? Perhaps it will mean the service will hire civilian coders or train airmen to write code.

We’ll see. Roper made clear he’ll be making some decisions about the best way ahead quite soon.