No matter what reforms government promises, planning remains a complicated, time consuming and costly business.
A Canberra property consulting start-up wants to take the pain out of submitting development applications and help speed up the delivery of new housing in the process, with the first AI agent built exclusively for the ACT’s planning system.
Urban Intelligence has developed Daaisy – Development Assessment Artificial Intelligence System – to cut through the complexity that slows development approvals, giving planners, developers and property professionals instant answers, grounded in the ACT Government’s planning rules and technical regulations.
It says talking to Daaisy would be like talking to a planner.
Urban Intelligences uses the power of the large language models such as ChatGPT and Gemini but its AI products like Daaisy are specific to the ACT.
Director strategy and analytics Adrian Makeham-Kirchner said that while general purpose models such as ChatGPT would provide answers that could be wrong, Daaisy relied only on the ACT Government documents for dealing with the statutory planning system and the Urban Intelligence’s proprietary geospatial data set.
“It only talks to that [ACT] body of knowledge that we’ve given it permission to engage with,” he said.
“It gives you a chance to query information specific to what you need in the planning space without having to worry about some of the hallucinations and junk that you get from the bigger models.”
This allows users to get down into the planning weeds and ask specific questions about individual blocks not just general information.
It will even provide a reference for the answer at the end of the query, saying which document – government guidance notes, legislation, district strategies, policies or technical specifications – it sourced.
Urban Intelligence has also captured all the information in the government’s mapping service, ACTmpi, and integrated it into its body of knowledge, which givers users access to the characteristics of the geospatial database, something other AI products cannot do.
A technology called Retrieval Augmented Generation also contextualises what would otherwise be a massive pile of data.
“It links the answer to the context of the question rather than just to the most likely word,” Mr Makeham-Kirchner said
“RAG allows for the large language models to ask is this in the context of a site question, is it in the context of a policy question, is it in the context of a rule and then it’ll give you the answer based on that context.”
Daaisy is now public and available to try for free as it continually updates every fortnight on a three-month course to a fully commercial launch.
It will be one of three platforms Urban Intelligence hopes to have tested and ready. The others are DA Assist for submitting development applications and a mapping service, Site Scout.
There has been interest from both the private and government sector but also some fear that these tools could take over from statutory planning processes.
Mr Makeham-Kirchner believes it won’t come to that but it could relieve planners and applicants of the more boring and perfunctory aspects of the process.
“I don’t think there’s any chance of ever replacing planners,” said. “Planning is just too complex but it would hopefully free up time and energy for more constructive relationships between planners and people submitting their DAs.”
One area where it could be particularly useful is expediting the smaller-scaled, less complicated but much-needed Missing Middle development.
Eventually the company wants to develop a national body of planning knowledge that could be localised depending on where one lived.
A national model would have a drop down in the interface for a particular state and then local government area.
“It’s not different to what we’re doing here, it’s just there are eight different planning systems, with 500 odd different sub-planning systems, and that the additional complexity outside of the ACT,” Mr Makeham-Kirchner said.
Mr Makeham-Kirchner joined ACT government officials and industry specialists at recent AI event hosted by the Property Council.
They explored how AI could help improve development approval timeframes in the ACT.
Property Council ACT & Capital Region executive director Ashlee Berry said approval timeframes remained a key issue for Canberra’s development industry, with delays affecting feasibility, delivery and overall confidence.
She said the ACT’s planning system was still too slow and uneven to deliver a steady pipeline of new homes.
In February, it took on average 74 days to make a decision, according to reporting from the City and Environment Directorate.
“Lengthy and uncertain timeframes increase holding costs, undermine feasibility and delay delivery, particularly for well-located multiunit projects,” Ms Berry said.
“AI has the potential to support better outcomes by reducing duplication, improving consistency and helping planning authorities focus resources where they add the most value – shortening the path from application to approval without compromising governance or quality.”