AI agents in real estate are moving into the back office. Who holds the keys?


The 2026 Proptech Forum spent a day on trust in an AI world as AI agents start quietly plugging straight into the systems that run property.
By Mark Hollands
Nobody at the Proptech Forum in Sydney this week needed convincing that AI has arrived. The theme was "From Adoption to Advantage. Blueprint for Trust in an AI World." The real question at The Collider venue was not whether to use AI but when AI stops helping us and starts deciding for us.
The plumbing is already going in. In September, Sydney construction software company Build Paperless listed its job management platform in Anthropic's Claude directory, giving AI agents 37 tools to read jobs, log site diaries and draft variations. Less than a fortnight later, PropertyMe, whose software runs nearly two million properties for 6,500 agencies, launched AiMe Connect, which lets an agency's own ChatGPT or Claude account query its live rent roll. In October, REA Group told investors it is "building the agentic platform for the property industry".
Most of us use generative AI, which writes and summarises when asked. Agentic AI is the next step, in which software takes actions inside other systems.
Beyond that phase comes artificial general intelligence, or AGI, which could emulate the quality of a skilled person. Then comes the claimed era of the scariest scenario – artificial superintelligence (ASI), which would do it better than any of us.
Possible scenarios
For now, in this latest round of announcements, these AI agents mostly read. PropertyMe's first version is read-only. That caution will not last. Here’s how it might play out.
In 2027, property management goes first. AI agents inside rent rolls will approve routine repairs under a spending limit, choose the tradie, chase arrears and issue standard notices without asking. Property managers will handle only the exceptions, and most agencies will not notice the day it happened.
Sales activities will follow maybe the following year, beginning in the back office and then moving to the pointy end. AI agents will qualify buyers, run inspection calendars, write the campaign and recommend the price guide from live data. Sellers will start asking why they pay full commission for work a machine did.
Perhaps inevitably, buyers bring their own AI agents. Let’s say they start doing that in 2029. A buyer's AI agent will shortlist homes, book inspections and make opening offers to the vendor's AI agent. Australia will see its first home sale negotiated machine to machine, and the portal search box will start to look old-fashioned.
It feels crazy writing the previous paragraph, but the speed at which innovation is moving makes this possible. Only the speed of adoption would make many of us feel such a scenario was improbable. Yet, the world is full of early adopters.
If we dare look into the next decade, commercial leasing and construction follow with AI agents managing leases, tenders and site programs. If AGI arrives at any time in the 2030 decade, one system could run most of a property manager's job end to end. If ASI follows, it could price, plan and finance housing.
So, if this is our possible future, why would property management be the first domino? Because it is rule-bound, high volume and already digital, and it is where the software makers have opened the door. Sales runs on relationships and emotion, so it moves second.
Who owns the customer?
That changes the economics of proptech. If the first screen a property manager opens each morning becomes ChatGPT or Claude, the property platform becomes plumbing. Valuable plumbing, but plumbing. In America, where strategy is a little ahead of much of the world, Zillow, Redfin and Realtor.com have already launched apps inside ChatGPT.
My contrarian view is that the winners will not be the platforms with the most AI features. They will be the ones whose data is clean and trusted enough for someone else's AI to act on. Gartner predicts more than 40 per cent of agentic AI projects will be cancelled by the end of 2027.
Don't give away the fun part
In May, the Malaysian comedian Ronny Chieng, a University of Melbourne law graduate, gave Harvard's class of 2026 a blunt warning. People love bragging about using AI to write their speeches, he said, but "the creating is the fun part". The coming battle would be "mastery versus faking it".
An appraisal, a negotiation or a hard conversation with a landlord is where a property professional builds judgement. Hand all of it to an AI agent and we risk bringing up future generations of workers who can run software but cannot read a room.
Ref: ADV-1793941508-M22





Comments