Choosing AI as a real estate agent: two pipelines, one buying decision

An agency runs two pipelines at once, and they have opposite economics.

The buyer pipeline is high volume and fast. One listing generates dozens of enquiries, a large share of them arrive outside business hours, and a buyer who hears nothing back books an inspection with whoever answers first. The decision window is measured in days.

The listing pipeline runs the other way. A handful of owners a month, each worth a fee that dwarfs any single buyer interaction, and a decision that takes months. The owner who mentions at an open home that they might sell next spring is worth more than every buyer in the room, and nothing in your CRM will bring them back up in eleven months.

Most AI sold to real estate agencies is built for the first pipeline. Speed to lead, instant response, inspection bookings. That work is real and worth doing. The commission sits in the second one. This guide covers how to evaluate a tool against both: the five criteria that separate a good fit from a bad one, the five categories of options in the market, how to think about total cost, the questions worth asking on every demo, and how to run a 30-day pilot that produces evidence before you sign.

The five things that separate a good fit from a bad one

Every vendor demo looks good. The demo runs on one channel, clean data, and a lead who replies. These five criteria decide whether a tool holds up in a real agency.

1. It treats buyer enquiry and vendor identification as two different jobs

A buyer enquiry needs an answer in minutes and an inspection time. An owner enquiry needs qualification, an appraisal booking, and a cadence that survives a year of nothing happening. A tool that runs both through the same sequence will do one of them badly.

Ask what happens when someone calls about a listing and mentions they have a property to sell. The good answer describes two different paths from that single call.

2. It writes into the CRM your listings already live in

Your pipeline sits in Rex, VaultRE, AgentBox, MyDesktop, or Box+Dice. A tool that captures leads into its own dashboard has created a second source of truth and a data entry job for whoever is least busy. Ask which fields it writes to, in which system, and what happens when the CRM ships an update.

3. It covers the hours your agents are physically unavailable

Saturday between eleven and three is your highest enquiry volume and your lowest availability. Your agents are at open homes, in the car, or on the auction floor. Evenings and Sundays carry the owner enquiries, because people decide to sell after watching a neighbour's auction.

Coverage is the whole point. Ask to see what the agent does with a call that arrives at 2pm on Saturday and again at 9pm on Sunday.

4. It respects the Spam Act, the Do Not Call Register, and your agency obligations

Automated outbound to buyers and owners sits inside the Spam Act 2003 and, for voice, the Do Not Call Register. Consent, sender identification, and a working unsubscribe stay the agency's obligation when an AI agent sends the message. Your state's agency licensing rules and the underquoting provisions that apply to price representations do not change because software generated the text. A vendor who has never raised any of this in a demo has left the risk with you.

5. It attributes signed listings back to first contact

Twelve months in, someone will ask what the tool returned. The answer needs to be a number: listings signed, source, days from first contact to authority, and what those figures looked like beforehand. That reporting belongs in place from day one, because the baseline is only available before you switch anything on.

A tool that reports response times and message volume is measuring itself. A tool that reports signed authorities is measuring the thing you get paid on.

The five buckets of options in the market

Every product sold to real estate agencies falls into one of five categories. Each solves a real problem, and each has a ceiling that arrives at a predictable point.

The point tool

An auto-responder on portal enquiries, a bulk SMS tool, a calendar link. Cheap, quick to switch on, useful. The ceiling arrives the moment someone replies. A point tool sends and then stops, so a person has to notice and pick it up. The buyer who replies during Saturday inspections waits until Monday.

The single-task AI agent

A voice AI agent that answers the phone, or a chat widget that qualifies web enquiries. These work well on the job they were built for. The ceiling is the handoff. The call gets answered and the details captured, then the record needs to reach your CRM, book the inspection, and start the right cadence depending on whether that caller was a buyer or an owner. Each of those is a separate purchase, login, and integration that someone owns.

The after-hours answering service

A person taking your calls when your agents cannot. Genuinely good at judgement and warmth. The ceiling is arithmetic and context. One operator covers one call at a time and does not know your listings, so what you get back is a message rather than a booked inspection or a qualified owner. Auction weekends produce the volume that breaks this model.

The in-house build

Someone technical wires up n8n or Make, plugs in an LLM, and connects it to the CRM. This can work well, and the running cost looks like nothing. The ceiling is ownership. The build lives in one person's head, breaks when a portal changes its enquiry email format, and stops when that person leaves. Ask who fixes it at 6pm on a Friday in eighteen months.

The integrated platform

One layer that answers the enquiry, works out whether it is a buyer or an owner, books the inspection or the appraisal, writes to the CRM, runs the follow-up, holds the long cadence, and surfaces the agent at the right moment. The ceiling is setup effort and cost. It asks more of you at the start and earns that back when both pipelines run as one process. Briick for real estate agencies sits in this category, alongside several others worth looking at.

How to think about total cost

Vendor pricing is the smallest number in this decision. Four other costs decide whether it works.

The licence. Per seat, per workflow, per conversation, or a flat monthly fee. Ask which, and ask what the number does when a big campaign launches. Per-conversation pricing that looks cheap in a quiet month behaves differently across a spring selling season.

The integration. Connecting to Rex, VaultRE, AgentBox, MyDesktop, or Box+Dice costs something, on the invoice or on your operations manager. Ask what it costs and who owns the connection when the CRM updates.

The configuration. Somebody defines what a qualified buyer looks like for your agency, what counts as an owner signal worth escalating, what the agent says, and when it hands to a person. That work is measured in hours of your team's time, and it decides whether the tool sounds like your agency or like a vendor.

The ramp. Between signing and the first listing that came through the new process, you are paying for both. Four weeks is normal. Twelve weeks means the tool is complicated or the vendor is slow, and both are worth knowing beforehand.

The cost of the current process. Most agencies skip this one. Count the appraisals you ran last quarter and the authorities you signed from them. Count the buyer enquiries that never got a call back. Multiply the gap by your average fee. Any vendor conversation that starts without that number is about price rather than return.

The seven questions to ask on every vendor demo

These questions change what happens after you buy. Ask all seven of every vendor and write the answers down.

1. Which fields does it write to in my CRM, and who owns the integration when it breaks?

Listen for named systems and named fields. A vendor who integrates with everything is describing a webhook and a developer. A vendor who names the contact record, the enquiry source, and the appraisal date in Rex has done this before.

2. Show me one call that turns into two different paths.

A caller asks about a listing, then mentions they own a place across the suburb. Ask the vendor to demonstrate that call producing both an inspection booking and an appraisal opportunity in the CRM.

3. Show me a nurture sequence on day ninety, in a live account.

Short sequences demo well. The listing pipeline is won late. A vendor who can only show you the first ten days built the product for buyer enquiries.

4. Where is consent recorded, and how does an unsubscribe propagate?

The answer should cover every channel at once, immediately. When someone opts out of SMS and keeps receiving emails, the obligation sits with the agency.

5. What does my agent actually see when someone replies?

The good answer is a notification with the full conversation and a recommended next step, delivered where the agent already works. The poor answer is another dashboard to check daily.

6. Which parts run without a person, and exactly where is the line?

Every vendor has a line between what the AI agent handles and what a person handles. The ones worth buying describe theirs precisely and let you move it. Treat any claim that the AI handles everything as a reason to ask again.

7. What does the attribution report look like on the day I renew?

Ask to see the actual report: listings signed, source, and days from first contact to authority, compared against your baseline. Where that report is missing, the renewal conversation runs on feelings.

Where Briick fits, and the 30-day pilot

Briick sits in the integrated platform category. A Briick Workflow is the sequence your agency already runs, written down: enquiry answered, caller sorted into buyer or owner, inspection or appraisal booked, follow-up delivered, long-cycle owners held in cadence, and the agent surfaced when someone re-engages. AI agents execute the steps across voice, SMS, email, and WhatsApp, and write every outcome back into your CRM.

In practice that is several workflows running together: an AI agent that sends your team only the leads worth converting, a callback workflow that clears the weekend backlog before Monday morning, a pre-appraisal sequence that builds authority with the owner before your agent walks in, a post-appraisal sequence that alerts your agent the moment engagement spikes, and a briefing workflow that gives you a full contact brief before every buyer and vendor call.

The gap this closes first is usually the weekend one. We wrote it up in The open home aftermath: where most sales agents lose buyers between Saturday and Monday, which is worth reading before any vendor demo, because it tells you which leak to pilot against.

Once the workflows are live, you run them by talking to @Briicky, the AI Operator. Ask what happened with a buyer, change a qualifying rule, or pull the attribution report by voice, between inspections.

The 30-day pilot

Every vendor conversation should end in a pilot, structured the same way every time. Thirty days, one workflow, three phases.

Days 1 to 14: baseline. Change nothing. Measure enquiries received, how many got a response and how fast, inspections booked, appraisals run, and authorities signed. Two weeks of honest numbers is what makes the next four weeks meaningful, and this data disappears the moment you switch a tool on.

Days 15 to 28: parallel. The agent runs alongside your existing process on one workflow, usually weekend enquiry handling, because it has the clearest baseline and the most obvious gap. Your agents keep working as normal. Compare the agent's output against the human process on the same leads and correct the rules that produce bad output.

Days 29 to 30: decide. Two weeks of baseline and two weeks of parallel running on the same pipeline makes the decision arithmetic. Response coverage, inspections booked, appraisals booked, and authorities signed, measured against the fortnight before you started. A vendor who resists this structure is telling you something.

FAQ

Which real estate CRMs does Briick write to?

Briick reads and writes to Rex, VaultRE, AgentBox, MyDesktop, and Box+Dice, and connects to most others by API or webhook. Enquiries from realestate.com.au, Domain, your website forms, phone, and SMS feed the same pipeline, so the contact record has one source of truth and nobody re-keys anything.

Will buyers and owners know they are talking to an AI agent?

Yes. The agent identifies itself as an AI assistant. Most buyers would rather get an inspection time at 9pm on a Sunday than wait until Monday for a person. When an agent should take over, the handover carries the full conversation, so nobody repeats themselves.

How does this handle the Spam Act and the Do Not Call Register?

Consent is recorded against the contact record and checked before every send. Unsubscribes are honoured immediately across voice, SMS, email, and WhatsApp at once. Every message is logged with a timestamp, so a full contact history can be produced in under a minute if a complaint arrives.

What happens to an owner who is a year away from listing?

They stay in a long nurture cadence with suburb sales data and check-ins paced to their timeline, and your agent is surfaced the moment that owner re-engages. This is where most of the value sits, and it is why a pilot should measure appraisals booked as well as listings signed.

Does an AI agent write listing copy or price properties?

Briick automates the workflow around enquiries, appraisals, and follow-up. Listing copy and pricing sit with your agents, where the professional obligation belongs. Any tool that offers to generate price guidance should be treated carefully given the underquoting rules that apply in most states.

How long until it is live, and what does it cost?

Both depend on your agency: which CRM you run, how your qualifying criteria are documented, and how many workflows you start with. Most agencies are live on their first workflow inside four weeks. Pricing is scoped after a short call and mapped to your enquiry volume, so you know the number before you commit.

See it run on your own pipeline

Bring last quarter's enquiry numbers and the authorities you signed. We will map the gap to the workflow worth piloting first and structure the thirty days so the evidence lands before you commit.

Book a Briick demo for your sales team.