Most of this software is not exciting or wildly innovative but it will make a big difference to people’s lives: ensuring people get the hospital appointments they want, that they are billed correctly for their banking services, that they don't wait in all day for the washing machine repairman.

Which brings me back to my lengthy preamble - in order to address this challenge what we need is not new things but improved processes. We need to be able to build better software, more cheaply, more repeatably and with less skilled people. We need to be able to find errors before they are frozen into code and we need to reduce the heavy cost of testing and retesting when changes are made. We need to design our software in ways that make it more robust when its operational environment changes, or when it requires enhancement. We would like to be able to predict and control 'feature interaction', that is emergent behaviour when a piece of software is composed with other software. We would also like procedures that allow us to determine in advance how much building a piece of software will cost us.

New processes are not sexy, even when they require new science they don't attract the spotlight. They are often introduced incrementally and their effects are only seen by producers. Consumers see new processes indirectly and in the long-term.