The state of California just announced that the new technology underpinning its Child Welfare System [pdf] won't be the usual "IT Solution" bought up in one big lump to follow a 4000-page specification. Instead, it's going to be built as a series of smaller modular projects driven by user needs, drawing on agile methodologies, a wider range of vendors, and, wherever possible, open standards and open source software. The decision, made in collaboration with Code for Americaand the federal government, sets an important precedent for how governments on all levels can get past the pitfalls of the standard procurement model.
Large institutional computer systems, especially in government, have long been their own punchline: expensive, bloated, malfunctioning, often designed around the needs of the small group of contractors qualified to bid for them, not their intended users. If visiting the DMV is bad for your mental health, working there and having to use its broken computer systems every day is worse. The list of recent Big IT failures covers not just Healthcare.gov's first iteration, or the broken health exchanges, but state welfare systems, city financial systems and regional emergency phone systems. The federal government has just frozen its chunk of paymentstowards Texas's new child support IT system until the contractors explain how it's going to get finished.
It's not easy to break the cycle of bad systems that emerge from bad procurement models. Choices are often constrained by legislation, influenced by lobbyists, or simply hemmed in by convention: no one ever got fired for buying IBM. It requires alternative approaches that work. Over the past five years, the UK's GDS has helped set an example of how to do things differently, and influenced the formation of similar in-house digital government teams around the world, including its American equivalent. Every time government chooses to do IT differently and succeeds, it redefines too-big-to-fail monolithic projects as risky choices, not comfortable or conventional ones. Or as CfA's (and MeFi's own) Dan Hon puts it: "It isn't hard to do this kind of thing. It isn't easy, either. It's just simply doable. The fact that you're not doing it is now a valid signal that you're not doing it for a reason."