This answer applies 3 methods from the guide, led by one.
Start with the repeated problem
Measure the task before selecting a tool, then choose the smallest intervention that could help.
- 1Describe the repeated task and who owns it.
- 2Measure time, missed work, errors and useful outcomes.
- 3Check fit, total effort, exportability, permissions and support.
- 4Prefer a reminder or template before a larger automation.
Hold when: Do not buy complexity to solve an undefined problem. Do not assume a tool is connected because it can chat.
Read it in Broker Brain →Keep one trustworthy working record
Use the CRM for confirmed identifiers, permission, needs, source dates and next actions.
Read it in Broker Brain →Test with a human checkpoint
Use permitted data, verify the output and keep every external action behind an accountable person.
Read it in Broker Brain →Start with the job, not the tool list
The guide's ‘Start with the repeated problem’ method asks one question before any purchase: which task do you do every week, who owns it, and what goes wrong with it now? For most agents the honest list is short: replying to portal enquiries, writing listings, following up after inspections and appraisals, vendor reports, and keeping the database alive between campaigns.
Note roughly how long each takes today and what slips. That is your baseline. Some of it needs no AI at all: a missed follow-up is often fixed by a reminder in the CRM you already pay for, not a new subscription.
General assistants: ChatGPT, Claude, Gemini and Copilot
These are the tools most agents start with, and they are genuinely good at words: rewriting a listing you have drafted, turning call notes into next actions, explaining a contract term in plain English for a first-home buyer, or rehearsing an appraisal conversation with a vendor who wants a higher number.
Their limit is that they are not connected to your business. Each new chat knows nothing about the buyer from Saturday's open home or what you promised them, and a figure they produce, such as a recent sale in the street, can sound right with nothing behind it. Use them for drafting and thinking, give them the facts, and keep client names and numbers out of personal accounts. The OAIC's guidance is plain: do not enter personal information, and particularly sensitive information, into publicly available generative AI tools.
The AI inside your CRM
Australian CRMs are adding AI that works from your own records, which is the thing a general assistant cannot do. Rex says its AI Assist lets you dictate notes, log calls, add contacts, set reminders and draft emails in one prompt, with a review before anything is sent, and that AI Prospecting gives a ranked daily call list with talking points; it lists AI Nurture, AI Manage and AI Automations as coming late 2026. MRI Vault says its CRM can generate property descriptions and sync listings to the portals from one place.
Those are the vendors' own descriptions, so test them on your work before you rely on them. If you use another CRM, ask the provider what its AI does today, not what is on the roadmap, and whether it is included in your plan or costs extra.
Listing, photo and design tools
Listing copy generators, inside your CRM or standalone, write quickly and fill gaps with details that sound right: a school catchment, a strata levy, a view. Your name is on the listing, so give them only checked facts. Photo editing, virtual staging and floor plan services such as BoxBrownie (which presents its work as done by professional editors rather than AI), and design tools such as Canva, save hours on campaign material.
Edited and staged images carry a rule of their own: the ACCC says agents must give truthful information about a property's features, so a staged or retouched photo must not mislead. Label staging clearly and follow your state's rules.
Five questions to ask any AI tool before you use it on client work
Ask these of every tool, including the free ones. A vendor who cannot answer them plainly has answered them.
- 01Where does client data go? Where is it stored, is it sent overseas, and is it used to train the vendor's models?
- 02Does it know my records, or do I paste the background in every time?
- 03Who checks the figures? Does it show where each number came from, and does it flag what it could not source?
- 04Does anything go out without me? Every message, listing and post should wait for a person to approve it.
- 05Can I get my data out, and what does it cost in total, including the time I spend correcting it?
Test one tool on one job for two weeks
The guide's ‘Test with a human checkpoint’ method keeps this cheap. Pick one job, try it first on a past listing or a made-up client, decide in advance what counts as a failure (an unsourced figure, a wrong fact, a line you would never say), and read every output before a client sees it.
After two weeks, compare the time the job took including your corrections, the mistakes you caught and what clients said. Then keep it, change how you use it or stop. That comparison, not a top-ten list, is the honest answer to which tool is best for you.
What gets in the way
Treating a ChatGPT subscription as having AI handled.
Keep it for drafting, and be clear what it cannot do: know your database, remember yesterday's follow-up or source a sale price. Those jobs need a tool that works from your records.
Adding a new app for every job and logging into none of them.
Check what your CRM's AI already does before you buy anything, and add a tool only when it takes a measured job off your plate.
Pasting vendor and buyer details into a personal AI account.
Describe clients by situation, not identity, follow your agency's data rules and the OAIC's guidance, and keep client records in systems your agency controls.
Keeping AI away from client work entirely because it might invent something.
You do not have to trust it; you have to be able to check it. Give it the facts, check each figure against its source and send only what you would sign.
Questions agents ask next
Is ChatGPT or Claude better for real estate agents in Australia?
Both write well, and the gap shifts with each release. What matters more is what you give it: the listing, the checked figures and the client's situation, without the client's name. Either writes a better draft from good facts than from a clever prompt with none.
Do Australian real estate CRMs have AI built in?
Several now do. Rex describes AI that logs notes, drafts emails and ranks a daily call list, and MRI Vault describes AI-generated property descriptions. Ask your own provider what is live today and what it costs, then test it on one job.
Can I put client details into ChatGPT?
The OAIC recommends not entering personal information, especially sensitive information, into publicly available generative AI tools. Use a tool your agency has approved for client data, and check your agency's privacy policy covers it. This is general information, not legal advice.
Can AI write my listings for realestate.com.au and Domain?
It can draft them, and your CRM may do it in one click. Give it only checked facts, read every line, and make sure the price and claims follow your state's rules and the ban on misleading advertising before it goes live.
What is the cheapest way to start with AI as an agent?
Pick one job you repeat every week, use the assistant or CRM feature you already have on it for two weeks, and measure the time including your corrections. Spend money only once you know which part of the job a tool has to solve.