In this G+T Snapshot series, our Banking and Projects and Charities teams will focus on the key issues – both legal and commercial – facing the social and affordable housing sector in Australia. Over the next six months, our team will publish a series of articles addressing topics such as the direction of and influences on, Australia’s social and affordable housing sector (including the impact of government and private sector involvement as well as the role and responsibilities of advisors, developers and social housing providers). We will also discuss the unique funding and infrastructure challenges faced by investors and financiers.
In our first edition, we consider one of the most fundamental issues facing the sector: where to locate social and affordable housing hubs and how housing providers, builders, funders and governments at all levels can understand and overcome the challenges faced in this selection process. The answer may come from a source we are all becoming increasingly familiar with: artificial intelligence (AI).
If the Federal Government’s ambitious plan to deliver 20,000 new social and 10,000 new affordable homes across Australia over a five-year period (including various targets set by the States and Territories) is to succeed, a number of challenges must be recognised, addressed and overcome. These include proactive community consultation, land costs, environmental issues, local planning and access to transportation, schools and support services. Central to all of these issues are two fundamental questions: where to locate social and community housing hubs and how to integrate them into existing infrastructure (or, if no existing infrastructure exists, how to best build and supply that infrastructure).
So how can AI be one solution to these challenges? What can a technology best known for generating song lyrics based on a user’s prompt, or creating images and sound possibly contribute to the social and affordable housing debate?
The answer lies in AI’s core functionality: data analysis and prediction and in the speed at which it can conduct this analysis and make these predictions based on an evolving database. There is no shortage of data on social (and suburb) demographics, land values, use and availability, population and migration growth trends, construction costs and local government planning and approval policies. All of this information is critical to selecting the most appropriate sites to locate new social or affordable housing hubs.
Faced with such a large volume of data, it is understandable why the individuals involved in the planning and site selection process (even the most experienced) are forced to either narrow their focus to just a particular data set or make general conclusions from a broad, but ultimately often incomplete, data set. It is also understandable why such an approach might not result in the best outcome in terms of deciding where, when and in what form a new social or affordable housing hub might begin.
AI could assist in the initial feasibility analysis by processing and analysing all that data and efficiently extracting key insights and trends that would otherwise be undiscoverable. For example, AI may be able to propose a combination of hub locations (inner city vs outer suburbs), development type (high density vs medium density), infrastructure services (close to existing rail networks vs scope for future public transport network expansion), demographic factors (an area already experiencing high development and high population growth vs an underdeveloped area) and land value (the cost of purchasing and redeveloping a brownfield site vs the cost of purchasing a greenfield site) that suggest the most optimal, or at least a useful shortlist, of potential hub locations. If requirements of government, social housing providers, investors or other stakeholders change, AI can re-analyse the data and apply it to the new parameters. This could potentially avoid material sunk costs and reduce upfront outlays when conducting the initial feasibility analysis.
While we see an emerging role for AI in this selection process, we also acknowledge that AI has its limitations and will not be the sole go-to source for every feasibility assessment. For example, AI may not be suitable for selecting the type of eligible residents for a development, as AI models have repeatedly demonstrated they can be equally as biased as any human, largely due to the inherent biases rooted in the data on which they have been trained.
Careful and selective use of AI can contribute to better planning decisions when it comes to social and affordable housing and we are excited about the opportunities it presents.
In our next snapshot, we will discuss the role of superannuation funds and other non-bank institutional capital providers in supporting the social and affordable housing sector and the challenges and opportunities this presents for the market.