AI on the Record · evidence register

Real estate · Lead follow-up and CRM · Record set RE-LF

AI Lead Follow-Up in Real Estate: Two Agents' Setups

Updated 4 recordsevery claim cites a recordresults self-reported

As of October 2026, our 4 cases show agents connecting an AI assistant to the CRM they already use, in every case Follow Up Boss. The assistant reads the notes, ranks who to call today and adds new contacts. One agent says he cut his CRM from about 1,200 contacts to 402 real people (self-reported). The evidence is thin: two people, and both sell AI training. The agent still makes every call.

Contents
  1. 1How agents use an AI assistant with Follow Up Boss
  2. 2What AI follow-up results they report
  3. 3Where AI lead follow-up goes wrong
  4. 4What to take from these cases
  5. 5Questions agents ask
  6. 6All cases for this task
  7. 7Records at a glance
  8. 8Related tasks

Keep one fact in mind as you read. All 4 records come from 2 people, and both of them sell AI training or a paid community (Aidan Holt, 2 cases; William Zhang, 2 cases). Most of the setups are shown on screen, but we have no account yet from an agent who has nothing to sell.

How agents use an AI assistant with Follow Up Boss

Both agents put an AI assistant for real estate leads on top of the CRM they already use. They do not replace Follow Up Boss. The assistant reads it and writes to it for them.

Who to follow up with today

William Zhang built an assistant on Claude and gave it the Follow Up Boss API documents. His morning routine works like this:

  1. Say “good morning”. William types it each morning to start the run.
  2. Get today’s list. The assistant reads the latest CRM activity and tells him which leads to follow up with that day (William Zhang).
  3. Get scores, drafts and tasks. It scores leads with a framework he adapted from Keller Williams and tags them A to D. Then it drafts a text for each lead and puts a task on his calendar.

In a later video he asks “who in my CRM should I follow up with today?” and gets hot, warm and housekeeping lists (William Zhang).

CRM clean-up and other jobs

  • Contacts added by voice. Aidan Holt built his own assistant, which he calls Atlas, but he does not name the model behind it. When another agent phoned him about a listing, he sent Atlas a screenshot of the call log and dictated one instruction. It added the caller to Follow Up Boss, tagged him and made a follow-up task (Aidan Holt). William adds leads the same way, from a screenshot of a text thread or by speaking the details (William Zhang).
  • A clean-up of the whole database. Aidan asked his assistant to go through every contact and write down what each client is looking for, so he can match clients to new listings (Aidan Holt).
  • Personal follow-up sequences. William had Claude rewrite his generic follow-up sequences to include his own YouTube videos, for example for Facebook leads who want to downsize (William Zhang).
  • Sells AI trainingYouTube2026-05-13

    William Zhang — lead follow-up and crm

    This is William's review after two months with a Claude-based assistant.

    Claude + Follow Up Boss · One flagged lead turned into a deal (no number)

What AI follow-up results they report

Two records give a number. Both are self-reported and not verified.

Reported results (lead follow-up and crm, self-reported)

Every number is the person's own claim in the video and is not verified. Bars show only records that state a time.

Aidan says his contact list went from about 1,200 entries to 402 “real” people, each with notes on what they want (Aidan Holt). That is a count of contacts, not a time saving or a deal.

He also says 15–20 seconds of dictation saves about 5–10 minutes of typing per contact (Aidan Holt). The video title promises “5+ hours a day”, but the video does not show that number.

William gives no number. He says one lead the assistant flagged became a hot prospect and then a closed deal, and that he would have missed it without the reminder (William Zhang). He also says his personal sequences get “a lot more engagement” than generic drips (William Zhang). One deal and one impression are not a measured result.

Where AI lead follow-up goes wrong

Most problems the agents mention are about time and attention, not wrong output:

  • The AI makes the list, but you still do the work. William still opens Follow Up Boss to act on what the assistant finds, and he copies and sends the texts himself (William Zhang). He also notes that Follow Up Boss already has the same activity data. The assistant only sorts it his way (William Zhang).
  • Tracking everything can burn you out. At first William had a daily report of every task he did not do. He says it made him feel guilty and “really burned me out”, so he cut it down to one main task a day (William Zhang).
  • No check on what gets deleted. The clip of the 1,200-to-402 clean-up does not show how Aidan checked which contacts were dropped (Aidan Holt). A contact removed by mistake is a lead you do not call.
  • Building the tools costs time. Aidan says building AI tools pulled him away from his important tasks for two weeks (Aidan Holt). He says his automation setup “took me a long time”, and in the same video a posting automation fails because a caption was too long (Aidan Holt).
  • Plans are not results. William describes an AI voice caller for new leads, but it is still a plan (William Zhang).

What to take from these cases

These four records support a small set of points. They do not show that AI follow-up works for most agents.

  • Start from the CRM you have. In every record the assistant works on top of Follow Up Boss, and nobody moved to a new CRM (Aidan Holt, 2 cases; William Zhang, 2 cases).
  • Keep the human step. Both agents still look at the result. Aidan looks over the finished CRM entry, and William sends every text and makes every call himself (Aidan Holt; William Zhang).
  • Check before you delete. Review a clean-up list before contacts leave your database (Aidan Holt).
  • Ask for less. One main task a day worked better for William than a full list of what he missed (William Zhang).
  • Count the setup time. Both setups are custom builds, and Aidan reports weeks of distraction while he built his (Aidan Holt).
  • Weigh the source. Both people sell AI training, a paid community or a setup service, so their videos are also sales pitches.

Conflict of interest, 4 records (lead follow-up and crm)

  1. Sells a course, kit or service4 of 4
  2. none found in the video0 of 4
Sells something: Aidan Holt, William Zhang. "None found" means we found nothing in that video, not that the person has no business interest. How we flag →

Questions agents ask

Can AI do my real estate lead follow-up?

It can make the list and draft the texts, but in these cases the agent still does the follow-up. William Zhang’s assistant ranks leads and drafts a text for each one, and he sends the texts and makes the calls himself (William Zhang, 2 cases). He sells AI training.

Does AI work with Follow Up Boss?

Yes, in all four records here. William connected his Claude assistant through the Follow Up Boss API, and Aidan Holt’s assistant adds contacts and tasks to it (Aidan Holt, 2 cases; William Zhang, 2 cases). Both agents sell AI training or a setup service, so weigh their claims.

Is anyone using AI for lead gen in real estate?

Not on this page. No lead follow-up case here finds new leads. The closest is an old lead from more than a year ago that the assistant marked as hot because she wrote to him again (William Zhang). Two other records use an AI voice agent to call leads; one reports 11 calls and no appointments after about a week (self-reported) (Jonathan Dickey; @justbigrealty).

See every record in the case register and the method in How we collect cases.

6. All cases for this task

Each card opens the full record: video, workflow, result, where it goes wrong, conflict of interest and source. Filter across tasks in the case register.

Current workflows (2025–26)

  • Shares numbersSells a paid communityYouTube2026-05-15

    Aidan Holt

    Another agent phoned Aidan about one of his listings.

    Own AI "Atlas" + voice dictation · 15–20 s of dictation instead of ~5–10 min of typing per contact · self-reported

  • Shares numbersSells a paid communityYouTube2026-04-24

    Aidan Holt

    Aidan asked his AI assistant to go through every contact in his Follow Up Boss CRM and make a list of what each client is looking for.

    Own AI assistant + Follow Up Boss · CRM cut from ~1,200 contacts to 402 real people · self-reported

  • Sells training + setup serviceYouTube2026-06-26

    William Zhang

    William says he dislikes working inside Follow Up Boss, so he lets his Claude assistant do it.

    Claude + Follow Up Boss · No number; old lead resurfaced as hot

  • Sells AI trainingYouTube2026-05-13

    William Zhang

    This is William's review after two months with a Claude-based assistant.

    Claude + Follow Up Boss · One flagged lead turned into a deal (no number)

7. The records at a glance

Shares numbers the person states a result number (self-reported)Sells … the person sells AI training, a kit or a serviceHow we order cases →

Cases that share numbers and don't sell anything come first; then cases with numbers; then the rest, newest first in each group. Negative, caution and earlier-approach records follow the current workflows.
Who · toolResult (self-reported)LabelsPublished
Aidan HoltAgent, RE/MAX MetropolisOwn AI "Atlas" + voice dictation · YouTube15–20 s of dictation instead of ~5–10 min of typing per contactShares numbersSells a paid community2026-05-15
Aidan HoltAgent, RE/MAXOwn AI assistant + Follow Up Boss · YouTubeCRM cut from ~1,200 contacts to 402 real peopleShares numbersSells a paid community2026-04-24
William ZhangAgent in Austin, TexasClaude + Follow Up Boss · YouTubeNo number; old lead resurfaced as hotSells training + setup service2026-06-26
William ZhangAgent in Austin, TexasClaude + Follow Up Boss · YouTubeOne flagged lead turned into a deal (no number)Sells AI training2026-05-13

Self-reported numbers, the order of cases and the "Sells …" label are explained once in How we collect cases.