A centralised platform approach for AI innovation needed in government

“Scalable, reusable, and interoperable,” are the terms used by the Singapore government to describe its technology stack, a platform through which common digital services and infrastructure are made available to all government agencies.

“As technologies like IoT, edge computing, and AI become increasingly adopted as smart city initiatives, the concept of platformatisation becomes even more critical – as the way to standardise security, stability and scalability,” said Red Hat’s Director of Solutions Architecture, Sivaram Shunmugam, in a webinar.

He was speaking at the Build Once, Deploy Anywhere: The Future of Whole-of-Government Analytics & AI panel discussion hosted by GovInsider, Red Hat, and SAS.

Other speakers included SAS’ Principal Cloud Architect, Abhilash P.A., GovTech Singapore’s Product Manager of Analytics.gov, Jeffrey Chai, and Australia’s Department of Social Services (DSS)’ Director, Data Access Branch, Paul Oakes. The panel was moderated by AI Singapore’s Director, AI Innovation, Laurence Liew.

Consistency and cost-savings

Several speakers highlighted that a centralised platform to deploy AI and analytics models can ensure consistency and provide potential cost savings.

SAS’ P.A. pointed out that it can take months for a developer to build an application. Without using a common platform, there is room for inconsistency across applications and a higher chance for human errors, he explained.

GovTech Singapore’s Chai shared the same opinion that a centralised platform eliminates the need for different agencies to build their own systems and eliminates costs.

Rather than individual agencies acting on their own, the pooling of resources can enable economies of scale and better results.

“We can aggregate the demands from the agencies, and resources to be utilised on the platform, such as GPU instances.

“Because we can commit to a certain level of utility, it’s very easy for a big platform like [GovTech Singapore's whole-of-government platform] to commit to saving plans from cloud service providers.

“Eventually we can translate all these cost savings back to the agencies as well as the end users,” he explained.

GovTech Singapore has developed a whole-of-government data exploitation platform to support analytics and machine learning projects by public agencies.

Cross-agency collaborations and citizen-centric services

A key driver of utilising a centralised platform is to enable citizens to interact seamlessly with the government, said DSS’ Oakes.

Citing the example of First Nations people in Australia, he underlined that a centralised approach to AI and analytics can make policy more inclusive.

“You have government agencies focused on policy and other agencies focused on delivery. But they may not be the same agencies.”

Having a centralised platform also opens opportunities for different agencies to work on common use cases and complex issues spanning housing, education, finances and more, said GovTech Singapore’s Chai.

At the same time, it also accelerates AI innovation across the entire public sector as the platform provides standardised scripts, best practices, and templates that the agencies can adopt right away for their workflows.

Making it easier for agencies to customise their own user apps benefits the last mile of public service delivery as well, adds Red Hat’s Shunmugam.

Addressing data privacy and security concerns

In the next few years, agencies need to treat data literacy as another key language for everyone to pick up to better appreciate the power of AI and ethics, said GovTech Singapore’s Chai.

Both Chai and SAS’ P.A. highlighted that when personal data is used for analytics purposes, they need to be masked, anonymised or even aggregated. Chai shares that GovTech Singapore’s data science team currently embeds data anonymisation into their due diligence process.

P.A. shared that SAS utilises a synthetic data maker that generates synthetic data that mirrors real-world data and better protects sensitive information.

“The security of the data must be encrypted at the start and while in transit. And your platform must be able to handle the data which is encrypted,” he adds.

As governments move from legacy systems to the cloud, they face the challenge of integrating across different data sources and formats. This becomes a problem for deploying AI and analytics, as high-quality data is the foundation of AI access.

He recommends a hybrid cloud model, instead of an on-premises model, for governments to keep their options open, as it can accommodate a wide variety of storage network servers and data formats, as well as to enable users to tap on a wider range of AI and analytics solutions that are available.

You can click here to watch the full webinar on-demand. Other topics including overcoming cultural inertia within government organisations, governing Generative AI and related legal liabilities were discussed as well.