Many councils in Australia support open data sharing of planning, environmental, asset and other datasets, so they can be repurposed and used to build models and applications that contribute to smart city initiatives.
This hands-on experience with structured data preparation and quality gives councils an edge as they move into more advanced data use cases, harnessing AI and machine learning models.
However, while it has positioned them well for the AI boom, it doesn’t change the fact that they will still encounter some challenges in the fast-growing AI space. Addressing these challenges can help local governments make the most of AI.
Balance investments in different types of AI
Generative AI has changed the way that organisations consider AI’s potential. It has sparked use cases across industries, leading to productivity improvements. Another great thing that GenAI has done is exposed work that many organisations were already doing with AI, but away from the spotlight. Without pressure to work with one type of AI over others, this behind-the-scenes work has typically taken a balanced view around what type of AI or machine learning to apply to a particular problem or scenario.
That open-mindedness can benefit broader AI adoption discussions and initiatives in councils today. Councils should harness current enthusiasm levels and educate themselves on how to use GenAI, but the result of any project doesn’t have to be built with GenAI. A force multiplier is being able to lead a rational internal conversation about addressing business problems with AI, where the right type of AI is adopted at the right time and for the right purpose.
Balance the challenge of the business problem
GenAI, and the large language models that underpin it, is well-suited to summarising huge volumes of text or documents into something more readable and easily understood. But not every problem in local government is this big of a data problem. In our experience, organisations often fall into the trap of asking GenAI to find answers in relatively small repositories of documents or data.
GenAI is an over-engineered solution to this kind of data problem, where the volumes are so small that someone with a search engine and a smartphone can be quicker and more accurate than a GenAI-powered chatbot. And so, the second thing councils should consider is the scale and complexity of the problem they’re trying to solve, versus the reward or impact of applying GenAI to it. There may be more cost-effective ways to address the challenge than with GenAI.
The risk of getting things wrong
Councils are custodians of a large volume of sensitive and valuable data. While this presents latent opportunities for raw data transformation and ingestion into AI and ML models, there needs to be consideration around the risk of those models producing erroneous outputs or forming a data-driven basis for bad decision making. One way of balancing this is by injecting a human into the decision-making loop, ensuring there is adequate oversight of the AI, its output, and its role in decision making.
Data is one foundational element
In the current environment, we’re constantly reminded that data is the foundation of all AI initiatives – and it is, but it’s not the only foundational element that needs to be considered. To generate any value from AI, you need to be able to integrate and optimise your AI models, data and infrastructure. Yes, the infrastructure and models will sit idle if you don’t have data, but it’s not just about getting data right.
Organisations often have the AI model and the data, but not the right infrastructure in place to enable computationally intensive AI workloads to scale. Leaders in the space understand they need to take a holistic view of AI’s foundations and evolve these foundations with their needs and sophistication over time to ensure their AI project is delivering optimally.
Taking a whole-of-council perspective
A key opportunity around local government data is being able to bring multiple datasets and streams from across council operations together to create value. That naturally lends itself to a whole-of-council approach, both to data and to the adoption of AI. From a data perspective, a key decision to be made is how to bring data together in one place for consumption – by other teams and by AI models.
While one approach is to set up an entirely new repository – such as a central data warehouse or data lake on-prem or in the cloud – it can be expensive to house all this data separately from the source systems, with the assumption that all or most of it will get re-used. An alternative cost-effective approach is data virtualisation. Data remains in its source systems but is centrally viewable and searchable. A team then fetches only the data it needs to power an AI model or capability, enabling them to move quickly while avoiding costly infrastructure.
A second reason why a whole-of-council model may be favoured is that it enables a platform discussion to take place. Councils need to ensure that they standardise on an AI platform that meets the needs of the entire organisation. This avoids the potential for duplicated systems and costs to materialise as AI becomes a core part of their operations.
Petar Bielovich is director – data & analytics, Atturra
Leave a Reply