Upskilling and embedding GenAI in existing workflows key priorities for public sector
By SAS
Establishing a clear link between GenAI and real-world use cases, as well as capacity building efforts are key to tackle adoption issues in the public sector, said speakers from the Australia and India governments, as well as analytics firm SAS, in a recent webinar.
Establishing a clear link between GenAI and real-world use cases, as well as capacity building efforts are key to tackle adoption issues in the public sector, says government and SAS speakers in a recent webinar. Image: GovInsider
The American government uses AI to process feedback from veterans, model soil moisture to improve drought and flood forecasting, and to help the Patent and Trademark Office search for past patents quicker – at least according to ai.gov, a website that requires US government agencies to declare their AI applications.
This repository was shared by global AI and analytics firm SAS’ Asia-Pacific Lead for Public Sector Consulting, Ensley Tan, at a recent webinar titled “GenAI-Powered Government – The Next Frontier of Public Services?” alongside speakers from the governments of Australia and India.
However, AI models should be employed for a clear purpose, and not simply for novelty, cautioned Tan, adding that citizens or public sector users working on the ground do not care about technology beyond how much concrete help it can provide them with.
In his keynote, he shared various generative AI (GenAI) applications in the government, ranging from chatbots for citizen queries, internal database search engines used by civil servants, and code-assistants for developers.
To subscribe to the GovInsider bulletin click here.
Sandboxes and public-private partnerships
How can agencies get started with their GenAI pilots?
For a start, Tan proposed that public sector organisations consider using sandboxes for employees to test solutions to real-world problems in a confined space.
“Sandboxes encourage [public sector] users to go there [to start using GenAI], discover what they know they can do, as well as discover risks that they would never be able to theorise.
“You only find out when real-world evidence comes about,” he explained.
The government of Telangana, India’s IT Advisor, Sai Krishna, shared that his state adopts a sandbox approach with GenAI adoption in the public sector.
Through the Telangana AI mission, the state government currently works with private sector innovators to work on different AI applications across different sectors, including agriculture, public infrastructure, social welfare, urban planning, and more.
He highlighted the importance of public-private partnerships: “We have about 125 companies that are working on use cases in the government. And private use cases actually help advance our general adoption [of GenAI] in the state,” he shared.
Telangana is one of the leading state governments in India when it comes to AI adoption, having deployed over 50 applications over the past four years, according to Entrepreneur.
Upskilling and raising awareness
Both government speakers, including Telangana’s Sai and NSW Department of Customer Service’s Director of Digital Strategy, Investment & Architecture Digital NSW’s Daniel Roelink, highlighted the challenge of AI literacy and upskilling efforts.
NSW has been trying to demystify AI, said Roelink, explaining that AI can come across as too technical.
He highlighted the critical role played by governments in setting the standards and guidelines around accountable and responsible use of AI.
The government has to be responsible for the data it stewards and remain cautious of potential issues, particularly when it concerns citizens privacy and intellectual property rights, he explained.
When it comes to regulating any emerging technology, he said agencies need to balance between raising users’ capability while supporting responsible use of technology in the government, he noted.
To subscribe to the GovInsider bulletin click here.
Solving the physical infrastructure gap
Another priority for capacity building efforts is improving the physical infrastructure. For Telangana, the challenges come in the form of ensuring internet connectivity for its approximate 10 million out of 40 million who live in the rural areas of the state, said Krishna.
Roelink also highlighted the need for data centre capability and infrastructure to run large-language models.
Australia has traditionally been a larger consumer of overseas LLMs, he said, but it is planning to build additional data centre capacity to deal with potential future demand.
Another option is for the government to tap on smaller models which are less expensive to train and run, he shared.
“Small language models, RAG powered solutions and other new approaches will just continue to drive costs down for GenAI, and also open more specific markets and use cases. So, I think in that sense, it's quite promising for more niche markets,” said SAS’ Tan.
“We’re all trying to find answers along the way and these answers are developing as we discover what’s going on [in AI],” he added.