A brokerage can use AI to draft a short four-touch outreach sequence to a list of agents you chose from public information, then to turn its onboarding steps into a dated checklist and a first-30-days plan. AI writes the drafts. A broker reviews every message, and nobody is promised income.
Most recruiting emails fail because they sound like every other recruiting email: split percentages, "tools and support", a link to book a call. AI makes that failure faster, because it will happily produce 200 identical messages by lunchtime. The advantage comes from feeding it specifics about your office and the person you are writing to, then keeping the volume low enough that a human still reads each reply.
Build the target list from public facts you chose
Start with a spreadsheet, not a prompt. Open a sheet with these columns: name, current brokerage, area they work, price band, a public fact (a recent listing, a community they specialise in, a class they teach), source of that fact, and a "do not contact" flag. Fill it from public sources such as brokerage team pages, your local board's member directory where it permits this use, and the agent's own website. Do not scrape private profiles or buy a list you cannot trace.
Then decide who belongs on it before AI touches it. Write down three criteria in plain words. For a twelve-agent brokerage that sells mainly family homes in a mid-priced suburb, they might be "works in our area or the neighbouring ones", "closed at least eight sales in the last twelve months in the $250,000 to $600,000 band", and "has a niche we lack, such as relocation or probate". Those are facts you can check. They are also easy to defend if anyone asks why a particular agent was approached.
AI can help with the sorting. Paste the rows (names and public facts only, no phone numbers) into your assistant and ask it to flag which rows meet each criterion and to say which column it used. Spot-check ten rows by hand. If the assistant guessed a fact that is not in the sheet, tighten the instruction: "use only the columns provided; if a fact is missing, say 'not in sheet'."
One line on tools: a business-plan assistant is the safer place for this, because ChatGPT Business and Claude Team do not train on your content by default, whereas consumer plans leave a model-training setting for you to check yourself (on ChatGPT's personal plans it starts switched on). The ChatGPT plan guide for realtors covers the cost side.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowWrite your office's recruiting pitch in four sentences first
AI cannot invent what makes your brokerage worth joining, and it should not try. Before drafting anything, write four sentences yourself: who you help (the agent profile), what you actually provide (named things, not "support"), what it costs them (your fee structure in one line), and what a first conversation looks like (twenty minutes, no commitment, no pitch deck).
A filled-in version for the same illustrative twelve-agent office: "We are a twelve-agent independent brokerage in the family-home market. New and mid-career agents get a named mentor for the first ninety days, a CRM set up for them in the first week, and a weekly pricing meeting where we look at live listings. Our fee structure is a flat monthly desk fee plus a capped split, which I will send in writing. A first conversation is twenty minutes and I will tell you honestly if we are not the right fit."
Keep this paragraph in a document. It becomes the "facts" block in every prompt below, so the AI never has to guess at your offer. If the paragraph contains a number, check it against your current agent agreement before every send.
The four-touch sequence, with sample drafts
Four touches over about three weeks is plenty. More than that starts to feel like pursuit. Each touch has one job, and each should work on its own if the agent only reads that one.
| Touch | Day | Job | Channel | Length |
|---|---|---|---|---|
| 1 | 0 | Show you noticed something specific and ask one easy question | 60 to 90 words | |
| 2 | 5 | Share one useful thing with no ask (a local market note, a meeting invitation) | 50 to 80 words | |
| 3 | 12 | Offer the twenty-minute conversation in plain terms | Email or LinkedIn message | 70 to 100 words |
| 4 | 20 | Close the loop politely and leave the door open | 30 to 50 words |
Here is the prompt that produces the first touch. It carries your facts block and one row from the sheet, and it forbids the phrases that make agents delete recruiting mail.
You are drafting a first recruiting email for the recruiting broker of a small independent real estate brokerage.
OUR OFFER (use only these facts):
[paste your four-sentence pitch]
THE AGENT (use only these facts):
Name: [first name]
Public fact: [e.g. "closed 11 sales in the last 12 months, mostly 3-bed homes under $450,000, and runs a first-time buyer seminar"]
Source: [e.g. "their own website"]
Write an email of 60 to 90 words. Mention the public fact in the first sentence. Ask ONE easy question about how they handle something specific. Do not mention commission splits, income, "top producer", or "unlimited". Do not claim we know their current brokerage's terms. End with the recruiting broker's name, the office postal address and a line saying how to opt out of further messages.
Plain and warm, no exclamation marks.An illustrative output, and what to fix in it:
Hi [first name], I saw your first-time buyer seminar on your website, and eleven sales in a year in that price band is steady work. Quick question: when a first-time buyer's mortgage approval comes in lower than the plan, who helps you re-run the numbers with them? We have a weekly pricing and finance meeting at our brokerage and I wondered if that is something you already have covered. If it would help to compare notes, I am happy to chat for twenty minutes. If you would rather not hear from me, reply "no" and I will not write again. [Recruiting broker's name]
Two fixes before sending. "Steady work" is a judgement about someone's career, which reads as patronising; cut it. And "eleven sales" must match what you can actually see publicly, so check the source line. Edit the draft rather than re-prompting, because each re-roll of the model produces a new set of small invented details to check.
Touch 2: give something away
The second email should be useful even if the agent never replies. Pull from material you already own: a one-paragraph note on how inventory in a neighbourhood changed, the agenda of your Tuesday pricing meeting, a checklist your agents use for pre-listing appointments. Ask the AI to turn it into 50 to 80 words and to add "you are welcome to join the meeting as a guest, no obligation". If you have no such material, skip the touch rather than manufacturing it.
Touch 3: say what the call is
Agents are cautious about "a quick chat" because they have been burned by hour-long pitches. Describe the call exactly: "twenty minutes, I will ask about your pipeline and tell you what we do and do not offer, and you will leave with our fee sheet in writing." Add a booking link only if it takes one click. Ask the AI for two versions, one for email and one for a LinkedIn message under 60 words, and use the shorter version if the agent is active on LinkedIn.
Touch 4: the polite close
"I will stop emailing after this one. If your plans change, my number is below." That is the whole message. Touch 4 often draws replies precisely because it is the first message that asks for nothing.
What to say when replies come in
Replies need a human. Set a rule that anyone who answers, even with a "not now", gets a message from the recruiting broker within one working day. Use AI to draft those replies but start from the reply, not the sequence. Paste the agent's message and ask for a two-sentence response in the same tone.
Three cases come up again and again. A curious reply ("what are your splits?") deserves the fee sheet in writing and the offer of the call. A polite refusal ("happy where I am") deserves a thank-you and a calendar reminder for six months. A hostile reply ("take me off your list") deserves a one-line apology, immediate suppression in your sheet and no further contact. Put the suppression step in the spreadsheet itself so it cannot depend on someone's memory.
One twelve-agent office: 120 emails, one new agent
This is an illustration, not a benchmark. Say the twelve-agent brokerage builds a list of 120 agents over two evenings. It sends touch 1 to all 120 on a Tuesday. Over three weeks, suppose nine agents reply in some way: four say not now, two ask for the fee sheet, two book a call, one asks to be removed. After the calls, one agent visits the office, and one joins eight weeks later.
The sum: 120 emails, 9 replies (7.5 per cent), 2 calls, 1 joiner. Hours of human time: about 5 hours to build the list, 2 hours to review drafts, 3 hours for replies and 1 hour for each of two calls, so roughly 12 hours for one agent. That number matters more than any rate. If the new agent produces enough transactions to cover the desk fee and the mentor's time, the effort pays for itself. If it does not, change the message or the list before you change the volume.
What the AI saved here was mostly drafting and sorting time, not the conversations. Rewriting 120 emails by hand would take the best part of a day; reviewing 120 drafted from the sheet took about two hours. The recruiting broker's judgement is still the scarce part.
The onboarding checklist, dated from the day they say yes
Recruiting does not end when an agent signs. A new agent who sits unsupported for a week starts looking elsewhere, and then the whole sequence above was wasted. Ask AI to turn your own process into a dated checklist, then edit it. Give it your list of tasks in any order (messy bullet points are fine) and this instruction:
Turn these onboarding tasks into a dated checklist for a new agent joining our brokerage. Group by "before day 1", "day 1", "week 1", "weeks 2 to 4". For each task give an owner (recruiting broker, office manager, mentor, or the agent) and a "done when" line that someone could verify. Do not add tasks that are not in my list; if something looks missing, list it separately under "questions for me".
TASKS:
[paste]A filled-in extract of what you want back:
| When | Task | Owner | Done when |
|---|---|---|---|
| Before day 1 | Agent agreement signed and countersigned | Recruiting broker | Both signatures filed in the agent's folder |
| Before day 1 | Email address, CRM login and listing-system access requested | Office manager | Agent can log in to all three on day 1 |
| Day 1 | Meet the mentor; agree the first four weekly check-ins | Mentor | Four meetings in both calendars |
| Day 1 | Add current contacts to the CRM, with permission to contact recorded | Agent with office manager | Import checked on ten sample contacts |
| Week 1 | Walk through a live transaction file from open to close | Mentor | Agent can name every document in the order it is needed |
| Week 1 | Read the office wire-fraud and file-retention policies | Agent | Signed acknowledgement filed |
| Weeks 2 to 4 | First pre-listing appointment attended with the mentor | Mentor | Debrief notes saved in the CRM |
The "done when" column is the useful part. It turns "set up the CRM" into something a second person can verify, which matters when the office manager is on leave in the agent's first week. Your compliance file review process is a natural place to check these items; the steps in AI compliance file review for brokers show how a transaction-file check can be partly automated, and the same logic applies to an agent's start-up file.
The first 30 days, week by week
Week 1: remove friction
Everything in the first week should be about not letting the agent feel lost. Set up their accounts before they arrive. Introduce them to three people: the mentor, the office manager and one agent who joined within the last year. Give them a one-page "who to ask for what" sheet. AI can generate this from your staff list in a minute; you then check names and roles.
Week 2: a first conversation with a client
Aim for a real client conversation by the end of week two, even if it is only a mentor-led buyer consultation. Agents join to do business, and the ones who do none for a month start asking whether they chose well.
Week 3: look at their pipeline together
Ask the agent to bring a list of everyone they intend to contact in the next 60 days. Use the AI to draft first-touch messages for each person for the agent to edit, not to send automatically. This links to the touch-plan approach in the past-client sphere-of-influence plan, which suits a new agent who is moving a database across.
Week 4: the honest review
A 30-minute meeting with three questions: what has been harder than you expected, what have we promised that we have not yet delivered, and what is the next 60 days' goal? Write the answers into the agent's file. If you promised a tool and it has not arrived, say so and give a date.
Where AI helps the new agent's first month
The new agent gets some of the benefit too. Give them the office's approved prompt for listing descriptions, the template for a follow-up after an open house, and the rule on what never goes into a public assistant: client names with financial details, contract terms, offers in progress. The office-wide rules on that sit in brokerage rules for emails, disclosures and follow-ups; hand the agent that page in week one.
A short before and after shows why this matters. Before: a new agent pastes a buyer's pre-approval letter into a free assistant to "summarise the conditions". After: the office policy says summarise from a typed list of the conditions with the name and figures removed, and use the business-plan assistant. It costs two minutes and keeps financial documents out of a consumer tool.
Compliance limits that sit around every message
These are practical limits, not legal advice. Your broker's lawyer or your association's legal hotline decides what applies to you, and rules vary by area.
- No income promises. Never let a draft say or imply what an agent will earn. Ask the AI to check its own draft: "list any sentence that promises or implies earnings".
- Respect existing agreements. Agents may be bound by non-solicitation terms with their current brokerage, or have listings that belong to it. Your messages should never ask them to bring pending clients or listings, and your calls should not discuss them.
- Opt-out and sender details. Commercial email generally needs an accurate sender identity, a valid postal address and a clear way to opt out, and the rules usually make no exception for business-to-business mail, so recruiting emails to other agents count. Honour every opt-out immediately (some laws set a firm deadline, such as ten business days) and keep a suppression column in the sheet.
- Public information only. Use details the agent has published about their business. Do not infer personal facts, and never mention someone's family, health or age.
- No protected-trait filters. Sorting the list by age, gender, ethnicity, religion or anything similar is discriminatory even when an assistant does it quietly. Keep the criteria to business facts you wrote down in advance.
- Training content. Fair housing and similar rules still apply to what you teach new agents, so keep that module human-reviewed and dated.
Reviewing the sequence after four weeks
Review after four weeks. Four numbers are enough: emails sent, replies of any kind, calls booked and agents joined. Put them in a small table beside the date and the version of the message used. If replies are under about 3 per cent, the first sentence of touch 1 is usually the problem; rewrite it using a different public fact before you touch anything else. If replies are fine and calls are few, touch 3 is probably vague about what the call is.
A realistic mistake looks like this. A recruiting assistant, working fast, sends touch 1 to a list where the "public fact" column was filled in by the AI, not by a person. Three emails mention a listing the agent never had. One of the agents replies, annoyed, and posts about it in a local agents' group. The fix is a rule: every fact in the sheet needs a source in the next column, and a human fills the source. It costs minutes per row and prevents the one kind of mistake you cannot undo.
When the sequence is running and a tool starts to matter, compare the CRM options before you buy anything: the realtor CRM comparison covers what a small office actually pays. For wider support that sits beyond recruiting, AI tools for real estate brokers maps the transaction and agent-support side.
If you would rather map your recruiting steps with someone, that is the kind of job my AI implementation consultation covers; you can reach me on the contact page.
Questions brokers ask before automating recruiting
Can AI write recruiting emails to agents at other brokerages?
Yes, as drafts. The rules that matter sit around the email: only approach people through public business details, include a working opt-out, respect any non-solicitation terms the agent signed, and promise nothing about income. Have your broker's lawyer read the first sequence once.
How many agents should I contact in a first batch?
Around 100 to 150 is enough to learn what works. A smaller list gives noisy results, and a bigger one sends a weak message to many people before you have fixed it. Review replies after two weeks, then adjust the message.
Should the AI decide which agents to recruit?
No. Use it to sort a list by public facts you chose, such as transaction volume in a price band you serve. The decision to invite someone to a call stays with the recruiting broker, and nothing should be filtered by age, gender, ethnicity or other protected traits.
What should an onboarding checklist include besides paperwork?
Paperwork is the easy part. The first 30 days are won by a named buddy, a first-week client conversation, CRM and email set-up done for them, and a Friday check-in. Agents who feel lost in week one tend to drift to another brokerage.
Further reads
- How to Automate Buyer and Vendor Follow-Up With AI — Follow-up automation for the clients your new agents bring.
- Follow Up Boss Pricing: What a Small Team Really Pays — What one popular agent CRM really costs a small team.
- Virtual Assistant or AI for a Real Estate Agent: What Goes Where — Decide which recruiting admin goes to a person.
- Email Newsletter QA Checklist Before You Hit Send — A send-time checklist you can reuse for each touch.
- AI Automation Agency for Real Estate: Scope, Fees and Red Flags — What outside help costs if you outsource the set-up.
- Rechat AI for Brokerages: What Lucy Does and What It Costs — One brokerage platform with built-in AI, priced.
- Claude or ChatGPT for Realtors: Which Assistant Fits Your Day? — Which assistant suits the drafting side of recruiting.
- AI for Staffing Companies: Client Onboarding and Job Order Intake — From intake call to structured job order and onboarding pack: prompts, a filled-in example and checks that stop a vague order costing placements.
- Caregiver Recruiting Strategies: AI for Ads, Screening and Follow-Up — Draft caregiver job ads, build fair screening questions and re-engage lapsed applicants with AI, with a human making every hiring decision.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT pricing page and Data controls FAQ; Claude pricing page and privacy centre; the FTC's commercial email compliance guide (sender identity, postal address, opt-out).