Real estate AI is most useful on the parts of an agent’s day that are fast, repetitive and easy to check: replying to portal enquiries, following up appraisals and open-home visitors, and drafting listing copy for a person to approve. It is risky anywhere it states a price, a fact about a property or a promise to a buyer or vendor without a person in the loop.
This article is for principals, sales agents and agency managers in Australia. Rules differ by state and territory, so check yours.
Where an agent’s time goes
Most agents do not lose business because they cannot sell. They lose it in the gaps: an enquiry that came in at 9pm on a portal and was answered the next afternoon, an appraisal that never got a second call, an open-home visitor who was never contacted, a vendor update that slipped a week. Meanwhile the agent is at a listing presentation or on a street doorknocking, which is where the money is earned.
Real estate AI earns its place by closing those gaps. It does not replace the relationship. It makes sure nobody waits.
Real estate AI for enquiry replies
A portal or website enquiry usually asks one of four things: is it still available, can I inspect, what is the price, can you send the contract. An AI assistant can reply within a minute, any hour, from your own listing information, and ask the one question that moves things on, such as whether the buyer is pre-approved or has a property to sell. It books the inspection into the agent’s calendar, adds the contact to the CRM with the property attached, and tells the agent.
What it must not do is quote a price, hint at what a vendor will accept, or promise a result. Those answers come from the agent. Set it to say “the agent will confirm pricing” and hand over. The ACCC states that agents must not mislead consumers about a property’s price or other information about it, and that this applies whether or not the misleading is deliberate (ACCC, real estate, opened November 2026). An AI that improvises a price is a quick way to breach that.
The speed argument is covered in how fast to reply to an enquiry, and phone calls are a separate job, covered in AI receptionists for real estate agencies.
Appraisal and open-home follow-up
This is where AI produces the most money for the least risk. Appraisals that do not list today are often listed later, with whoever stayed in touch. A simple sequence does the staying in touch:
- A thank-you the same day, from the agent’s name, with what was discussed.
- A message a few weeks later with a relevant local sale, chosen or approved by the agent.
- A check-in at the vendor’s likely decision point, such as before spring.
- A flag to the agent when the person replies, opens a message or clicks a link.
For open homes, visitors who signed in can receive a follow-up the same evening with the property details, similar listings and an easy way to reply. The AI personalises the wording; the CRM keeps the sequence going; the agent takes the replies that matter. Our guide to real estate automation covers the follow-up machinery. This is the same idea as display home visitor follow-up in the building trade.
Listing copy with a person checking
AI writes a decent first draft of a property description from the agent’s notes and photos in a couple of minutes. It also adds things nobody told it: a “north-facing living area” that faces south, “walking distance to the station” that is a 25 minute walk, “recently renovated” for a bathroom from the nineties. These are typical invented details, and in advertising they matter.
Under the Australian Consumer Law, the ACCC page above says agents must give truthful and complete information about a property, including features, location, zoning, history, possible land use and price. So the rule is simple: the agent verifies every factual claim in AI-written copy against the property before it is published, and strips the adjectives that cannot be proven. Treat the AI draft as a junior’s draft.
A checklist for AI-drafted listings
- Every measurement, room count and distance checked against the plans or the property itself.
- Every claim about condition, renovation, aspect and views confirmed.
- Zoning, approvals and land use claims verified, not assumed.
- No price indication unless it matches a documented appraisal or the vendor’s acceptable price.
- Photos and descriptions still match each other.
On price specifically, the ACCC says agents should advertise a price based on a reasonable market appraisal or what the seller will accept, and should not advertise well below that price to attract interest. Several states also have their own underquoting laws, and Fair Trading or Consumer Affairs bodies enforce them. Check the rules in your state.
What the agent still does
| Task | AI | Agent |
|---|---|---|
| First reply to an enquiry | Replies, asks a qualifying question, books | Takes the conversation from there |
| Appraisal follow-up | Keeps the sequence going | Calls when the contact responds |
| Price guidance | None | All of it |
| Listing copy | Drafts | Verifies every fact, approves |
| Vendor updates | Summarises activity into a draft | Adds judgement and sends |
| Negotiation | None | All of it |
Privacy and records
Agents hold names, phone numbers, finances and sometimes ID. Keep enquiry data in your CRM and use AI tools that sit inside it or under a business agreement, and do not paste buyer or vendor details into free consumer chat tools. The Australian Privacy Principles considerations are in our AI and the Privacy Act guide. Tell people in your privacy policy that you use automated tools, and let them reach a person quickly. Commercial messages need consent and an unsubscribe; for the detail see SMS marketing rules in Australia.
Making sure the AI does not make things up
An assistant answering buyer questions from a listing should only use the listing data you give it, and should say “I will check with the agent” otherwise. This is a set-up decision, explained in how to reduce AI hallucinations. Test it before it goes live by asking about things the listing does not say, such as school zones or strata levies, and see whether it guesses.
A worked example: the evening enquiry
Imagine a portal enquiry arrives at 8.40pm on a Tuesday for a house the agent is showing on Saturday. In many agencies it waits until about 10am the next day, by which time the buyer has also written to three other agents. With an assistant in place, the buyer gets a reply within minutes confirming the inspection time, asking whether they are pre-approved, and offering to book a private inspection on Thursday evening. The agent sees a ready-made summary in the morning and a booking in the diary. Nothing about price was discussed, because the assistant is not allowed to.
This is an illustration, not a result. Whether it lifts your inspection numbers depends on your listings and your market, which is why you should measure replies, bookings and inspections before and after.
Vendor updates and seller communication
Vendors want to know what happened this week: enquiries, inspections, feedback, comparable sales. AI can pull the figures from your CRM and portal reports into a clear draft, group buyer feedback into themes and suggest talking points. The agent adds the judgement and the advice on price, and decides whether a hard conversation is better held by phone.
Two cautions apply. First, feedback summaries can smooth over a strong objection, so read the raw comments on a regular basis. Second, anything that suggests the property is attracting more interest than it is can mislead the vendor, so the numbers in the update must be real counts from your own records.
Where to start
- Check your current response time on portal enquiries, including evenings and weekends.
- Automate the first reply and booking before anything else.
- Add appraisal and open-home follow-up, with agent approval of anything that mentions price or market conditions.
- Use AI for listing drafts only with a verification step built into your workflow.
- Review a sample of conversations each fortnight.
If you want this built into your CRM so it books into the real calendars and writes to each contact, that is the kind of work CentriWeb does in CRM set-up for property businesses. If your CRM already has follow-up sequences you are not using, start there first.