What this automation does
Most teams do not lose leads to bad agents. They lose them to slow first replies and follow-up that stops after two attempts. This workflow handles the first conversation and the long nurture, and puts agents on the phone when it matters.
- Capture. A new lead arrives in Follow Up Boss from Zillow, Realtor.com, your IDX site, a Facebook lead form or an open house sign-in. A webhook fires the workflow.
- Check before texting. The workflow looks up whether the contact already exists, is assigned to an agent, is in an active transaction, or has opted out. It also pulls the listing they asked about from your MLS data feed and confirms it is still active.
- First reply. The AI drafts a short text that names the property and one useful fact (open house time, price change, a feature from the listing) and asks one question. The text goes out automatically during allowed hours, or waits for agent approval if you prefer.
- Qualify. As the lead replies, the AI works through a short checklist: buying or selling, timeline, financing status, whether they already have an agent, and what they want to see. Leads who say they are already represented get a polite close and no further outreach.
- Book. When the lead wants a showing or a call, the AI offers times from the agent’s calendar and creates a task in the CRM with a summary of the conversation.
- Hand off. The assigned agent gets an alert with the lead’s answers. The AI pauses as soon as any human on the team sends a message in that thread.
- Nurture. Long-timeline leads get a light, personalized check-in on a schedule, built from their saved searches and inquiry history, until they re-engage or opt out.
When it is worth building
Build this when you get enough leads that response times slip, especially evenings and weekends, or when leads go quiet after one or two attempts because agents are busy with active clients.
Do not build it if a solo agent handles a handful of leads a month. Your CRM’s action plans and a saved text template are enough. Some real estate CRMs and lead platforms now include AI texting features, and if yours covers your lead sources and lets you set guardrails, try it first. A custom build wins when you need to pull live listing data, apply your own qualification rules, route across several agents, and log everything back to one CRM.
Fair Housing and consent guardrails
The Fair Housing Act prohibits steering buyers toward or away from areas based on protected characteristics, and a chatty AI can drift there fast. Put these limits in code and in the prompt:
- No opinions about neighborhoods, safety, “good areas,” school quality or who lives somewhere. The AI points to objective sources and offers the agent.
- No questions about family status, religion, national origin or disability. The qualification checklist is about the home and the transaction, not the person.
- Every outbound message is logged in the CRM, so a broker can review what was said.
For texting, the TCPA applies. Text only leads who gave consent through the portal or your forms, never purchased lists, honor any opt-out request in plain words, and stay inside daytime hours in the lead’s time zone. Real estate is also subject to state rules on advertising and solicitation. This is a design checklist, not legal advice.
Architecture
| Layer | What we use | Why |
|---|---|---|
| Lead hub | Follow Up Boss | Webhooks on new leads and an open API for notes, tasks and assignment |
| Listing data | Your MLS’s RESO Web API or your IDX vendor’s feed | Messages must reference live status and accurate details, under your MLS’s data license |
| Messaging | Twilio, or the CRM’s built-in texting where it exposes an API | Two-way SMS with delivery status and opt-out handling |
| Conversation AI | Claude Sonnet 5.5 with a fixed qualification checklist | Natural replies that stay on script and return structured answers |
| Orchestration | n8n, calling Follow Up Boss through the HTTP Request node | Handles timing, quiet hours, pauses and retries in one place |
| Scheduling | Google Calendar or Calendly for agent availability | Offers real open times instead of “when works for you?” |
How to build it
- List your lead sources and current rules. Who gets which leads, how fast, and what counts as a qualified lead for your team.
- Connect the CRM webhook. Start the workflow on new leads, and store the CRM contact ID with every message.
- Connect listing data. Map the inquiry’s MLS number or address to the listing record, and refuse to reference any listing that is no longer active.
- Write the first-reply prompt. One fact, one question, under 300 characters, signed as your team’s assistant. Have agents rewrite the examples until they sound like your team.
- Build the qualification loop. Structured output for each answer, a confidence flag, and a hard stop when the lead says they have an agent or asks for a human.
- Add guardrails. A blocked-topics list with neutral replies, the pause-on-human rule, opt-out handling and quiet hours.
- Add booking and handoff. Calendar lookup, CRM task with summary, and an alert to the assigned agent.
- Run in approval mode for two weeks. Agents approve each AI text from their phone. Switch to automatic sending once the edits stop.
The downloadable starter template below covers the webhook, listing lookup, first-reply draft and CRM note for Follow Up Boss and Twilio.
What breaks in production
- Duplicate leads. The same buyer inquires on Zillow and your IDX site an hour apart. Match on phone and email before texting, or they get two different conversations.
- Stale listings. The home went under contract after the inquiry. Check status at send time, and offer similar active listings instead.
- Agents texting in parallel. An agent replies from their personal phone, outside the CRM, and the AI keeps going. Make the CRM the only texting channel for leads in the workflow.
- Other agents posing as leads. “I’m the buyer’s agent” should route to the listing agent, not into qualification.
- Fake numbers. Portal forms collect typos and fake phones. Watch delivery failures and fall back to email.
- Prompt drift. The AI starts promising things, like “the seller will take less.” A banned-claims check runs on every outbound message.
Timeline and cost
Most teams are live in two to four weeks. Lumeor builds this in the range of $5,000 to $12,000. The low end is one CRM, one texting channel and first replies with handoff. The high end adds MLS data, multi-agent routing rules, calendar booking and long-term nurture. Running costs are SMS fees from your messaging provider (check current pricing) and a small amount of model usage per conversation. Our n8n messaging automations use the same orchestration pattern on other channels, and the Follow Up Boss tool page covers its API. For general B2B lead scoring, see the AI lead qualification agent guide.
How to measure success
- Median time from lead arrival to first reply, including nights and weekends
- Share of leads who reply at all
- Showings or appointments set per 100 leads
- Leads handed to agents, and how many agents rated as genuinely qualified
- Opt-out rate per conversation
- Closed transactions traced back to AI-started conversations