How Can Real Estate Agents Use AI for Follow Up?

What AI Follow Up Means in Real Estate

AI follow up refers to the use of generative AI and automation to draft, personalize, schedule or send the messages an agent uses to stay in contact with leads, past clients and the personal sphere after a first conversation. It typically covers text messages, emails and social messages, and in some products it extends to automated replies to new inquiries outside business hours. The underlying discipline is the same as conventional follow up: a planned sequence of contact with a reason for each touch, as described in How Should Real Estate Agents Follow Up With Leads?

AI changes the cost of each message rather than the purpose of follow up. Before generative tools, an agent either wrote each message by hand, which limited volume, or sent the same template to everyone, which limited relevance. Generative AI makes it possible to produce an individual draft for each contact in seconds, provided the tool has the facts needed to make the draft specific. AI follow up is distinct from initial lead response, which concerns how quickly an agent reaches a new inquiry and is covered in Why Does Response Time Matter for Real Estate Leads?

Most practitioners describe three levels of automation. In the first, AI drafts and the agent writes the final message. In the second, AI drafts and the agent approves or edits before sending. In the third, AI drafts and sends on a schedule without individual review. Agents commonly assign different message types to different levels based on the stakes of each message.

AI Adoption and Reported Impact Among Agents

Two industry surveys describe how widely AI is used. According to Delta Media Group's 2026 AI survey, as reported by WAV Group on January 30, 2026 and by HousingWire, 97% of brokerage leaders surveyed said their agents use AI, and only 4% said their brokerage is not using AI at all. Delta Media's survey reflects the views of brokerage leaders rather than agents themselves.

NAR's 2025 REALTORS Technology Survey asked agents directly. NAR conducted the survey in July 2025, inviting a random sample of 49,233 active REALTORS and receiving 1,241 usable responses, according to HousingWire. According to NAR and Florida REALTORS, 46% of respondents reported using AI-generated content, such as listing descriptions. On frequency, 20% said they use AI tools daily, 22% weekly, 27% a few times a month, and 32% had not yet used AI in their business. ChatGPT was the most used AI tool at 58%, followed by Gemini at 20% and Copilot at 15%.

Reported impact trails adoption. According to NAR and HousingWire, 17% of respondents said AI had a significantly positive impact on their business, 33% reported a moderately positive impact, and 46% said AI had no noticeable impact. The survey does not break impact down by how agents use AI, so it does not show whether follow up specifically performs better or worse than other uses. It does indicate that access to AI tools alone does not produce results for roughly half of agents who responded.

Established agents who want AI follow up built and run for them can review the Always-On Agent System at theinnercirql.com/agents/.

Why AI Follow Up Often Reads as Automated

Generic output is the most common complaint about AI-written messages. A general-purpose AI model asked to write a follow up text with no further context tends to produce the most typical version of such a message, which includes stock phrases such as "I hope this message finds you well" or "just checking in." Recipients recognize these phrases from bulk marketing, and a message that could have been sent to anyone is commonly read as if it was.

Four causes recur in practice. The first is the absence of a voice sample, so the AI has no model of how the agent actually writes. The second is the absence of personal facts, so the AI fills the message with pleasantries instead of content. The third is length, since general models tend toward multi-paragraph output while personal text messages are usually one or two lines. The fourth is sending without review, which allows errors such as a wrong name or an inappropriate congratulation to reach clients.

Each cause has a corresponding fix. Agents supply a set of their own past messages as examples, include specific facts from their records in each request, set explicit length limits, and route personal messages through an approval step. None of these fixes require a particular product; they apply equally to a general chatbot, a CRM with built-in AI or a done-for-you system.

The Approval Model: What AI Sends and What Agents Send

A common way to structure AI follow up is to sort messages by stakes. Low-stakes, recurring touches with a predictable shape, such as neighborhood market updates, home anniversary notes and periodic value content for the sphere, are commonly drafted from an agent-written template and sent on a schedule, with periodic spot-checks. Messages to a specific person about a specific matter, such as a reply to a question or a note to a buyer who has gone quiet, are commonly drafted by AI and approved or edited by the agent before sending. Sensitive conversations, including bad news, contract or money questions and anything emotional, are commonly handled by the agent directly, as are phone calls to people ready to transact.

The approval model reflects where repeat business comes from. According to the NAR 2025 Profile of Home Buyers and Sellers, as reported by NAR and nowbam, 91% of buyers would use their agent again or recommend them to others. Past clients therefore represent a large share of an agent's future business, and a single poorly judged automated message to a past client can carry more cost than a missed one.

The agent's sphere and past clients are usually organized as a separate list with its own schedule, as described in How Do Real Estate Agents Build a Referral Database? AI follow up tends to be most effective on this list when the database contains notes about each person that the AI can draw on.

Voice Matching and Personalization

Voice matching is the practice of giving an AI tool examples of the agent's own writing so that drafts resemble the agent's style. Practitioners commonly use 10 to 20 real messages that received replies, along with written preferences such as sentence length, phrases to avoid and how messages are signed. Most general AI tools allow these instructions to be saved so they apply to every draft without being repeated.

Personalization depends on facts the AI cannot produce on its own: the contact's name, property, timeline, family details the contact has shared, and the last subject discussed. These facts usually come from CRM notes, which makes note quality after each conversation a direct input to AI follow up quality months later. A request that includes these facts produces a message with a specific reason for contact, such as recent sales on the contact's street, rather than a general check-in.

Disclosure is a related practice. When an AI assistant conducts a live exchange with a consumer, such as answering an after-hours inquiry, practitioners commonly identify it as an assistant and hand the conversation to the agent promptly. Drafts that the agent reviews and sends under their own name are generally treated as the agent's own messages.

Rules on Text Opt-Outs and AI Voices

Consent rules apply to AI-sent messages in the same way they apply to messages an agent sends by hand. According to Brownstein Hyatt Farber Schreck and Day Pitney, the Federal Communications Commission's 2024 consent revocation order treats replies such as stop, quit, end, revoke, opt out, cancel or unsubscribe as a reasonable means of revoking consent, and requires callers and texters to honor revocation requests within a reasonable time not to exceed 10 business days. The FCC later delayed part of the order, and sources differ on the scope of that delay. Agents commonly configure AI tools to remove an opted-out contact from text follow up on the same day.

AI-generated voices are covered by federal robocall rules. According to the law firms Cooley and Wiley, the FCC unanimously adopted a declaratory ruling on February 8, 2024 confirming that AI-generated voices are "artificial" under the Telephone Consumer Protection Act. As a result, calls that use AI-generated voices are subject to the same requirements as other artificial or prerecorded voice calls, which generally include prior express consent. The ruling did not create a new category of rule; it clarified that existing rules apply. None of this constitutes legal advice, and agents typically consult their broker or an attorney before deploying AI voice calls.

Measuring AI Follow Up

Reply rate is the most direct measure of whether AI follow up is working, since it reflects whether recipients treat a message as personal. Open rates and send volume measure activity rather than response. Practitioners commonly review a small sample of sent messages each week, count genuine replies, and update the voice examples with messages that drew strong responses.

Results from follow up typically appear over months. According to the NAR 2025 Profile of Home Buyers and Sellers, as reported by NAR and nowbam, 91% of buyers would use their agent again or recommend them, which means the value of consistent contact with past clients accrues as referrals and repeat transactions rather than immediate closings. NAR's 2025 Technology Survey finding that 46% of agents saw no noticeable impact from AI suggests that impact should be measured against a defined use, such as sphere follow up, rather than AI use in general.

Frequently Asked Questions

Established real estate agents who want AI follow up that sounds like them, run on a system, can schedule a call with Blake Suddath and Tyler Lewis at theinnercirql.com/agents/.