AI Agents for Real Estate: 6 Workflows That Stay Governed
Real estate agents can use AI for lead follow-up, listing operations, transaction coordination, and property management. The durable version runs on governed apps, not a conversation that forgets.

Why real estate teams are evaluating AI agents
AI agents for real estate work best on coordination: qualifying inbound leads, matching buyers to listings, drafting listing copy, routing transaction documents, triaging maintenance requests, and assembling pipeline reports. The safest first workflow is lead intake and follow-up, because it repeats daily, touches several systems, and has a natural human checkpoint before anything goes out.
If you run brokerage operations, you already have a CRM, email, calendars, and a transaction tool. The gap is the connective work between them and a record of what happened. On Major, the enterprise platform where agents build the software they run on, an agent builds a stateful app for the repeatable parts. The app stores client and property history, runs routing and reminders the same way every time, and calls the agent only when a message needs judgment. Reason once. Run forever.
Scope note: this is workflow architecture, not legal, fair-housing, MLS, or brokerage advice. Check NAR's AI guidance and your own state and brokerage rules before you automate anything client-facing.
Six AI agent workflows for real estate
Each one splits into repeatable work (code in an app) and judgment (the agent or a broker).
1. Inquiry qualification and follow-up
Trigger: a new portal or web-form inquiry lands in the CRM. The app handles dedupe, owner assignment, consent state, and reminders. The agent reads the message, summarizes intent, and drafts a reply with a viewing slot. Human gate: the assigned agent approves every outbound message.
2. Buyer-to-listing matching
Trigger: a new listing or a changed buyer profile. The app filters listings against the criteria the buyer actually stated. The agent ranks the shortlist and explains the fit in plain language. Human gate: the agent reviews before anything is sent. The matching step never infers protected characteristics or steers by neighborhood demographics.
3. Listing copy and asset checklist
Trigger: a listing moves to "prep." The app tracks photos, disclosures, and required fields. The agent drafts descriptions from the facts in the record. Human gate: the listing agent verifies every factual claim before publish.
4. Transaction document routing
Trigger: a document arrives in the deal inbox. The app files it against the transaction, tracks deadlines, and flags missing items. The agent classifies ambiguous attachments and drafts the nudge to the other party. Human gate: the coordinator approves anything that changes a deal record or goes to a client.
5. Maintenance request triage
Trigger: a tenant message or work-order form. The app logs the request, applies urgency rules, and assigns a vendor from an approved list. The agent reads free-text descriptions and flags likely emergencies. Human gate: the property manager approves vendor dispatch above a spend limit you set.
6. Portfolio and pipeline updates
Trigger: a weekly schedule. The app pulls stage counts and stale deals from the CRM. The agent writes the narrative for the team lead. Human gate: reports go to internal channels only.
Where human review belongs
Drafting and routing automate more safely than negotiation, pricing advice, eligibility decisions, or financing and legal statements. Keep a person on all four.
Review points to build in:
- Any outbound message to a client or lead.
- Any claim about a property, such as square footage, taxes, school zones, or condition.
- Consent. The app stores consent state per contact and blocks outreach when it is missing or revoked.
- Fair housing. The agent never filters or ranks on protected classes, and a person reviews match logic before launch.
- Jurisdiction. Rules differ by state and brokerage, so compliance owns the review checklist, not the agent.
Because the review rules live in the app as code, they apply to every run. A prompt instruction can be skipped on a bad day. A required approval step cannot.
How to build a real estate lead follow-up agent on Major
This is a build specification, not a customer case study. The four pieces are the HubSpot, Gmail, Google Calendar, and Slack connectors plus one Major app.
- Trigger. A new HubSpot contact or inquiry record starts the workflow. A daily scheduled run checks for stale leads.
- State. A Major app with a managed database holds the inquiry source, preferred area, budget range as the lead provided it, property IDs, assigned agent, consent state, interaction history, next action, status, and an audit event for each step.
- Repeatable steps in code. Dedupe, owner routing, consent check, stale-lead timers, and status changes run in the app with no model call.
- Judgment in the agent. The agent reads the Gmail thread, judges intent, summarizes the request, proposes relevant listings, and drafts the reply. It reads available viewing slots from Google Calendar.
- Approval gate. The draft goes to the assigned agent in Slack. Gmail sends only after approval. Ambiguous or sensitive questions, such as financing or legal topics, escalate to a person instead of getting a drafted answer.
- Governance. Each connector gets only the read and write scopes it needs. Role-based access limits records by team. Every action lands in the audit trail.
You can start building free on Major and stand up this app against your own HubSpot and Gmail accounts.
A walkthrough, illustrative only. A buyer inquiry about a three-bedroom arrives. The app matches it to an existing contact, confirms consent, and assigns the agent on that territory. The agent summarizes the request, picks two listings that fit the stated budget, and proposes two viewing slots. The assigned agent sees the draft in Slack, edits one line, and approves. The app logs the send, sets the next action to a three-day check-in, and the stale-lead job picks it up if nobody replies.
The first time, the agent works out this flow. After that, the app runs it. That is the difference from a chatbot, which re-reasons each lead and keeps nothing. For more patterns, see AI agents for sales and AI agents for lead generation.
Start with one workflow and expand
Pick the first workflow by four criteria. It repeats daily. It touches two to four systems. A human can review the output in under a minute. A mistake is reversible. Lead follow-up meets all four. Document routing is a good second. Pricing and negotiation do not belong on the list.
Once the lead app exists, later workflows reuse it. Matching reads the same client and property records. Pipeline reports read the same status history. Each workflow the agent figures out becomes another governed app your whole team can run and oversee, which is how enterprise AI moves from pilots to governed work. The same pattern applies in AI agents for business workflows beyond real estate.
Some teams will never need this. If you send a handful of inquiries a week from one inbox, a template is enough. Once leads, listings, and deals span several tools and several people, durable state and approvals are what keep the work safe, and building them as an app is how Major makes that durable.
Start building on Major
Build a lead follow-up app that keeps client context, routing, and approvals in one governed workflow, while the agent handles judgment and the app runs repeatable steps.
Related articles
Frequently asked questions
- Which AI agent is best for real estate?
- Choose by criteria, not a vendor ranking. Look for multi-system workflows, durable client and property records, scoped permissions, and human review before anything is sent. Major fits that bar because its agents build deterministic apps that hold state, run routine steps the same way every time, and log every action.
- What can AI agents automate in real estate?
- Agents can handle lead follow-up, buyer-to-listing matching, listing copy drafts, transaction document routing, maintenance triage, and pipeline reporting. Routine coordination suits automation. Pricing, negotiation, and legal or financing advice stay with licensed professionals.
- Can an AI agent follow up with real estate leads?
- Yes. The agent summarizes the inquiry, drafts a reply, and suggests viewing slots, while an app checks consent, routes the lead, and logs history. The assigned agent approves each message before it sends, and ambiguous or sensitive questions escalate to a person.
- How should real estate teams govern AI agents?
- Grant each connector only the scopes it needs, restrict records by role, keep durable records and audit logs of every action, and require human approval for outbound messages and record changes. Compliance should own checks for fair housing, privacy, and local rules.