This is a slightly edited version of a piece I wrote for Policy Options, the magazine of the Canadian Institute for Research on Public Policy. Their full post is here.

It’s hard to say exactly when a government becomes “digital.” You could try looking at the UN’s e-Government survey, perhaps, or examining the health of your flagship technology projects. If you are a policymaker, you know it’s working when your job changes. And if your job isn’t changing, the chances are you are getting in the way of the development of digital government.

Many governments have tried to stop government IT projects from being a byword for disaster. Most have failed. The reason they failed was because they focused on the wrong things. Sometimes they focused on changing the IT, throwing more and different technology at the same problems. Sometimes they focused on changing the project  – asking suppliers to sprinkle some private sector magic onto it; or creating ‘innovation labs’ that, cordoned off from day-to-day business, allowed a few lucky public servants to work in new ways.

By bringing in new processes, culture, people and technologies, digital governments try to do things differently. Those changes range from creating blogs to communicate with people openly and establishing open source technology across government, to publishing real-time performance data for services, and changing recruitment processes. When it comes to digital, the trick isn’t to try and change the IT or the project. It is to change government.

Changing government sounds like a grandiose, if not impossible, ambition for any public servant. And maybe it is, if you’re trying to change everything at the same time. The key to making progress toward a digital government is to start small, to do the hard work to make service delivery simple, and to be ruthlessly focused on meeting users’ needs.

None of these things come naturally to any big organisation, let alone a government. Making them happen requires a small but meaningful shift in what policy means.

In most administrations built on the Westminster model, policy stands first among equals. Those who work in policy enjoy proximity to politicians and prestige. Traditionally, good policy requires two things: an understanding of citizens and an understanding of government.

Policy generalists are good at writing about large-scale complexity. But being good at writing about hard, practical things – designing a strategy to build a low-carbon economy, for example – isn’t the same as tackling those policy challenges. Working on paper gives policymakers a pass on thinking hard about implementation. Sometimes doing very prosaic things, like sketching out what a new public service could look like in code, showing webpages on a tablet to real people and then trying to improve that service on the basis of what they have said, can expose the cracks between the words.

To an experienced policymaker, this can seem facile. Many have spent years making assumptions about human behaviour. Some have tried their best to do their own user research. Having someone suggest that you should go and watch some human behaviour instead is insulting. There’s much more to policy, you might think, than just watching human behaviour.

But one of the things that we’ve found in GDS – and with other governments – are the ways the policymaker’s role needs to change. It no longer makes sense to have policymakers be the sole arbiters of how citizens and businesses might react to a minister’s promises to reframe the welfare system or improve public health. Distanced from the frontline, they are often no longer the best qualified for the task.

Digital governments are composed of multidisciplinary teams tasked with delivering public services. Policymakers are part of that mix, instead of calling the shots. The role of the policymaker in these teams is not to be the voice of the user; in a digital government, users use their own voices to give feedback. Instead, the role of the policymaker is to make sure that those voices can be heard inside government. They defend those voices when there are competing priorities, and make sure the minister is given the opportunity to hear them.

Policymakers are no longer only shaping delivery choices according to policy. They are also shaping policy choices according to delivery, drawing insights from tangible experiments rather than abstract theories. In the UK, there was a famous case of a service where a policy hinged on the legal definition of a couple. The minister responsible had his own, clear idea of what “a couple” meant. The service’s early users turned out to have very different, somewhat more flexible views on what it meant – views that might have been missed. The result of not heeding the users might not have been just political embarrassment; the unintended consequences of doing so might have been missed welfare payments, trips to foodbanks, even homelessness.

Giving ministers access to those voices matters. So does shaping political decision-making on the strength of the best available evidence. That is what policymakers do in digital governments.