Build it with AI
Build a real estate CRM around inquiries, properties, appointments, and follow-up.
Create a purpose-built relationship and opportunity system for agents, teams, brokers, or property businesses.
Introduction
What a real estate CRM has to hold.
Most teams end up with a real estate CRM the same way: a generic CRM bent around listings, plus a spreadsheet for the parts it cannot hold. A personal system works for an agent handling their own enquiries. It fails as soon as enquiries arrive across portals and cover is shared, because the first response wins and nobody can see which enquiries are unanswered.
A generic CRM models a deal as a single opportunity, which does not describe a transaction with two sides and a chain. There is no single queue across sources, no property context attached to the enquiry, and no record of response time — which in this market is the variable that most determines the outcome.
What follows covers building a real estate CRM: the records it holds (listings, buyers, sellers, viewings, offers, chain position, and next action), the systems it reads (Google Calendar and Gmail), and what it does not fix.
The problem
Enquiries from six portals and a response time that decides the outcome.
General CRMs model the contact and have no concept of a property, so the two most important facts about an enquiry live in different places. Portal inboxes model the enquiry and end at the notification.
The records are listings, buyers, sellers, viewings, offers, chain position, and next action, and the authoritative copy of most of them already lives in Google Calendar or Gmail. The portal shows twelve enquiries, the CRM has seven contacts, and the five missing were answered from a phone and never recorded.
The cost is not the inconvenience: a chain stalls because nobody was tracking the party who went quiet.
You're likely here because
- Nobody can say which of today’s enquiries have not been answered
- A generic CRM models a deal as a single opportunity, which does not describe a transaction with two sides and a chain.
- When it is wrong, a chain stalls because nobody was tracking the party who went quiet
What gets built
Launch builds it, Grow operates it.
Built in Launch
- • Lead records
- • Property context
- • Pipeline board
Operated through Grow
- • Lead response
- • Scheduling
- • Follow-up
Systems it reads
- • Google Calendar
- • Gmail
- • HubSpot
The record model
What an enquiry record has to hold.
- Enquiry, linked to person and property
- The enquiry is the unit of work. A system modelling only the contact loses the fact that decides how to respond, which is which property they asked about.
- Source portal
- Because response expectations and conversion differ by portal, and a blended figure hides which source is worth the fee.
- Response clock
- From arrival, with escalation. First response is the variable most correlated with winning the instruction in this market, and it is rarely measured.
- Owner assigned at capture
- So shared cover does not produce two agents contacting the same applicant, which costs the relationship immediately.
- Viewing and its outcome
- With a required follow-up step, because the gap after a viewing is where agencies quietly lose applicants who were genuinely interested.
- Applicant requirements
- What they are actually looking for, so a match six months later is a query rather than a memory.
- Consent and marketing basis
- Recorded at capture. Property enquiry data is routinely reused for marketing and the basis is usually assumed rather than recorded.
How it runs
From scattered enquiries to one responsive pipeline.
Step 01
Describe what a real estate CRM has to do
Model the enquiry as the unit of work, linked to both a person and a property. Systems that model only the person lose the reason the person got in touch.
Step 02
Connect the systems of record
Email carries portal enquiries, the calendar holds viewings, and existing systems supply property data. Reading email is what makes a single queue across sources possible at all.
Step 03
Build the operating surface
An enquiry queue across sources with response clocks, property context on each, a viewing scheduler, and follow-up that continues after the first appointment.
Step 04
Start narrow
One queue of unanswered enquiries across every source, with a response clock. Nothing else matters until that exists.
Step 05
Route the exceptions
An enquiry unanswered past the response threshold escalates to another agent, because an enquiry answered late by the right person converts worse than one answered quickly by anyone.
Step 06
Measure transactions with a dated next action against every active party
Measure time to first response at the median and ninetieth percentile, and viewings booked per hundred enquiries. Response time drives the second more than anything else does.
Implementation path
Building agent tooling that works from a phone.
- 01
Consolidate every enquiry source into one queue before improving anything else. Partial visibility produces confident management of the sources you can see.
- 02
Baseline response time at the median and the ninetieth percentile. The tail is where the lost instructions are, and it is invisible in an average.
- 03
Attach property context at the point of capture, so the first reply can be specific rather than generic. That specificity is most of the conversion difference.
- 04
Design for a phone first. Agents work from a car park, and a system that assumes a desk will not be updated until the evening, if at all.
- 05
Build the narrowest useful version first: the buyer and seller sides of one transaction on one record, with the next action visible to both agents.
- 06
Consolidating every enquiry source into one queue via email capture is the first week and is the prerequisite for everything else. The response clock with escalation is a further week. Design for a phone from the start — agents work from a car park, and a system assuming a desk gets updated in the evening if at all.
- 07
Once the response clock is live, add applicant requirements so a later match is a query rather than a recollection. Post-viewing follow-up as a required step comes next and closes the second-largest leak.
Controls
Controls that matter.
Control 01
A response clock per enquiry with escalation, since first response is the variable most correlated with the outcome in this market
Control 02
Consent and marketing-permission basis recorded per contact, because property enquiry data is routinely reused for marketing and the basis is often assumed
Control 03
Ownership assigned at capture so that shared cover does not produce two agents contacting the same applicant
Examples
Three enquiries that stop going cold.
The enquiry answered at 9pm
A single queue with a response clock and escalation means the enquiry is answered by whoever is available within the window, rather than waiting for the named agent to finish a viewing.
The applicant who came back six months later
Enquiry history against the person, including what they viewed and what they said, means the second conversation starts from the first rather than from nothing.
The viewing that produced no follow-up
Follow-up modelled as a required next step after a viewing closes the gap where most agencies quietly lose applicants who were genuinely interested.
How it goes wrong
Three ways agency systems fall behind.
Only the portals with structured integrations are in the queue, and the rest are managed from a personal inbox.
Capture from email so every source lands in one queue. Partial visibility produces confident management of the sources you can see and silence about the rest.
Response time is reported as a median and looks fine, while the enquiries being lost sit in the tail.
Measure the ninetieth percentile. The lost instructions are in the tail by definition, and the median is specifically the statistic that cannot see them.
Lettings is forced through the sales pipeline, and referencing and compliance steps have nowhere to live.
Model lettings states explicitly. The enquiry and viewing model is shared; what happens afterwards is genuinely different and does not fit a sales pipeline.
Limitations and considerations
What faster follow-up will not achieve.
- Faster follow-up does not create stock. In a supply-constrained market a better enquiry process changes which agent wins the instruction, which is worth having and is a different claim.
- Portal integrations vary in what they permit, and some provide notification only. Confirm what is available per portal before designing a queue that assumes structured data.
- Property enquiry data is personal data reused for marketing more often than the consent basis strictly supports. Record the basis at capture rather than reconstructing it later.
- If one agent handles their own enquiries end to end, a personal system works. In a supply-constrained market a better enquiry process changes which agent wins the instruction and does not create stock, which is worth being clear about internally.
- Connector coverage varies: Google Calendar, Gmail, HubSpot are representative rather than guaranteed, and the fields exposed depend on your workspace permissions.
FAQ
Build a real estate CRM with AI: common questions.
Why not use a general CRM?
Because the property is missing from the model. In this business the enquiry is about a specific property at a specific moment, and a CRM that holds only the contact loses the fact that decides how to respond.
How fast does first response need to be?
Faster than the other agents the applicant contacted, which in practice means minutes rather than hours for portal enquiries. Measuring the ninetieth percentile rather than the median is what exposes the ones being lost.
Can it pull enquiries from the portals automatically?
Email-based capture works across essentially all of them, which is why it is the reliable starting point. Structured integration is better where a portal offers it and should not be assumed to be available.
Does it work for lettings as well as sales?
The enquiry and viewing model is shared; what differs is what happens after — referencing, compliance checks, and renewals. Model those as their own states rather than forcing them into a sales pipeline.
What should the first version contain?
The buyer and seller sides of one transaction on one record, with the next action visible to both agents. Everything else waits until that one is genuinely used.
How will we know whether it worked?
Measure transactions with a dated next action against every active party against the baseline taken before anything changed.
Start with ARIA
Ask ARIA to build it.
Describe the website, application, workflow, or operating surface you need. ARIA plans, connects, builds, tests, and keeps refining it — inside the permissions you set.
- ARIA acts only through the systems and permissions you connect.
- Connections use scoped credentials you can change or revoke.
- Actions are recorded, and consequential ones can require approval.
Start here
Build a real estate CRM around the process you actually run.
Put every enquiry in one queue with a response clock, attach the property at capture, and design it for a phone.