The secret to using AI in government
By Cloudera
Four steps to get data ready for machine learning.
Contact tracers can use AI to comb through vast arrays of data sources; agencies can use it to minimise manual work and reallocate staff to vital new tasks like building an emergency quarantine facility; spending and welfare organisations can use it to target subsidies and find green shoots in the labour market.
But you are what you eat, and so is a data-crunching computer algorithm. Shaun Bierweiler, President, Cloudera Government Solutions, shared four steps to ensure that government datasets are ready for AI.
Preparing the ground
“Whether it be automated, augmented or truly artificial intelligence, there's tremendous potential for technology to assist and enable many agencies,” Bierweiler says. “But first and foremost, you have to ensure that you have the right data strategy.”
Policymakers must think outcomes first, technology second. Data strategy must not be “dictated by your application or legacy infrastructure decisions,” he warns. There are four essential steps agencies must take to ensure that they are ready for AI.
First, is the security of data. “Government systems and infrastructures are under constant threat or cyber targeting, and you need something that is able to deliver security and governance ”.
Second is openness. Governments should not be locked into a legacy system or one particular hosting service - even if a long term contract sounds too good to be true. “We believe very passionately that it needs to be compatible with any cloud, any hardware, any deployment platform”, he says. Cloudera’s commitment is to be as accessible as possible so that it can maximise the utility of data and work with anything.
This leads to a ‘hybrid cloud’ approach with data coming from multiple sources and being stored in different places. “Data across different agencies will naturally be generated in different applications stored in different clouds,” Bierweiler adds . “Having a data strategy that is independent and agnostic from where the application resides is key.”
The fourth is scalability. “Agencies must set out their strategy to enable this to happen”, he says. “Scale must be at the heart of their plans so that the system doesn't discriminate against your data and provides you with flexibility and freedom to innovate”.
Setting the strategy
The great military strategist Clausewitz once said that “the simplest thing is difficult” when talking about strategic leadership, and that is certainly true for data analytics. Agencies must ensure it is simple to maximise data use, but that requires forethought and strategic planning.
“Without the forethought of building in the necessary security governance at the onset, it's really difficult to go back and try to make the system government ready”, Bierweiler warns. “Put in the checks, protocols and security steps from the start rather than engineering them around an existing strategy.”
Interoperability is also vital. Can you use historical datasets? Will your data combine with sensors and IOT devices? This is particularly interesting as Covid-19 has led governments to reassess their data, and find new and urgent uses for it.
“Cloudera customers are finding really creative and innovative ways to use data”, Bierweiler says, “even tying them back to their historical data to enable things like predictive and proactive analytics”.
But take a step back, plan the data strategy, and ensure that it covers the key four pillars of security, interoperability, openness and scalability. The strategy shouldn’t simply be innovative, it should be solid. From there, great things will come.