A commercial broker can use AI to turn public-record owner data into a one-page profile, rank the owners most likely to act, and draft a short sequence of personal messages that refer to something specific about the property. The broker supplies and verifies the facts; AI organises, summarises and drafts, and you send each message.
The weak point in most owner outreach is not the writing, it is the research behind the first line. A message that says "I noticed you have owned the building since 2014 and the lease on the ground floor ends next year" gets read; "I specialise in your sector" does not. So most of the effort belongs before any drafting: public-record research, a one-page profile per owner and a check that every hook is true and current.
What counts as public-record research, and what does not
Stay inside sources you can name. For most brokers that means land or property records held by the local registry or assessor, company registry filings for owner entities, planning or permit databases, published accounts, press releases and the owner's own website. Your CRM and your own notes on past conversations also count, as they are yours.
What does not belong: data bought or scraped without checking its terms, private information on an individual that came from a source you cannot name, and anything a client told you in confidence about another party. If you cannot say where a fact came from, do not put it in a profile and do not use it in a message.
| Fact | Typical public source | Confidence |
|---|---|---|
| Registered owner name and entity type | Land or property record | High, if the record date is recent |
| Date of last transfer and price | Land record (price sometimes withheld) | High for date, check price |
| Other properties held by the same entity | Record search by owner name | Medium: similar names need checking |
| Directors, filings, registered address | Company registry | High for filings, may be out of date |
| Planned works or applications | Planning or permit database | High for what was filed |
| Tenant mix and lease ends | Site visit, listings, your own knowledge | Medium: confirm before quoting |
| Owner's intentions | None | Do not guess; ask them |
Say plainly in the profile which items are verified and which are inferred. That distinction keeps your first message honest, and it stops a plausible guess becoming a "fact" in a later email.
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 nowBuild the owner profile in one page
A profile is the unit of work. One page per owner, with the same fields each time, so you can compare owners and hand the research to a colleague. Collect the facts yourself from the records, paste them as plain notes (no personal data beyond what the public record shows), and ask AI to put them into the template and flag gaps.
Owner profile (one page)
Owner entity: [name, type, registered date]
Property: [address, type, size, tenants if known]
Held since: [date of last transfer]
Other holdings: [list, with source]
Possible triggers: [lease ends, refinance window, planning filing,
years held, recent vacancy]
What we do not know: [list]
Verified (V) / inferred (I) marks on each line
First-line hook: [one specific, true observation]
Next action: [who, what, by when]The "what we do not know" line is the one that stops research becoming fiction. If the answer is "tenant lease ends, who owns the neighbouring plot, whether they want to sell", then your message should ask a question, not state a belief.
A prompt that fills the template from your notes
You help a commercial real estate broker prepare owner profiles.
Use ONLY the notes below. Do not add facts. Fill the profile
template. Mark every line V (verified by a record I cite) or I
(inferred). List what is unknown. Suggest three possible
first-line hooks that are true on these facts alone. If the notes
do not support a hook, say so.
NOTES: [paste your notes with the record source for each]"Use ONLY the notes" and "do not add facts" matter. Without them the tool may fill gaps from general knowledge, such as typical lease lengths or the sort of business that "usually" occupies a building like this. Those guesses read as research and are not.
A worked example on a small multi-tenant building
Say a broker covers light industrial and small office buildings, and a target is a 14,000 sq ft multi-tenant building with six units. The broker's notes, from records and a site visit, are as follows (illustrative): owned by a private company registered 11 years ago, last transferred 9 years ago; the same company holds two other small buildings within a mile; planning database shows a roof-works application filed last spring; two of six units look vacant on a drive-by; listing sites show one unit advertised for 14 months.
The profile comes back with owner and property filled in, "held nine years" marked V, "two other buildings" marked V with the source, "two units vacant" marked I because it comes from a drive-by, and "unit advertised 14 months" marked V from the listing page. Unknowns: whether the owner is financing the roof works, whether they want to sell or lease, who actually decides. Three hooks: the roof application (true), the three-building portfolio (true), the long-advertised unit (true, but a touchy subject).
Now rank. A simple score helps you decide which owners to contact first, so the sums are visible rather than a hunch.
| Signal (illustrative scoring) | Points | This owner |
|---|---|---|
| Held 7 years or more | 2 | 2 |
| Holds three or more nearby assets | 2 | 2 |
| Capital works filed in last 12 months | 1 | 1 |
| Vacancy visible or advertised over 6 months | 2 | 2 |
| Prior contact with us | 1 | 0 |
| Total | 8 possible | 7 |
A 7 out of 8 goes to the top of this week's list. The scoring is a tiebreaker, not a prediction: the points are yours, and the weights should come from your own experience of which signals led to conversations. Review them after a quarter against what actually happened.
Write a first message that is specific and short
Give the tool the hook and the purpose, and tell it what to avoid. Messages under 90 words get read; messages that sound like a mail merge do not.
Draft a first email of under 90 words from a commercial broker to
the owner of a six-unit building. Use this one hook: the roof works
application filed last spring. Purpose: ask if they would like a
short, free view on how comparable buildings nearby are leasing, no
obligation. Do not state that I know their plans, do not mention
vacancy, do not use "I hope this finds you well", do not claim
results. End with a question they can answer in one line. Sign off
as [your name]. Subject line: under 6 words.Illustrative output: "Subject: Your roof works and nearby leasing. Hello [first name], I saw the roof works application for the building and wondered if you have looked at how comparable units nearby have been leasing this year. I have a short note on three recent deals and would be happy to send it over. Would it be useful? [your name]". What you fix: the tool may add "I specialise in..." or a claim about "great results". Remove them. Check that the hook is true, as a wrong hook on an owner who knows the facts ends the conversation at once.
A four-touch sequence, spaced and varied
A single message rarely lands. Plan four touches over about five weeks, each adding something different, and stop when the owner answers or says no.
- Day 0: the hook. The email above. One question.
- Day 8: something useful. A short comparable summary or a market observation, with the source. No ask beyond "tell me if this is useful".
- Day 21: a different channel or angle. A short letter or a call where your rules allow, referring to the first email. Mention a different true observation, such as the portfolio.
- Day 35: the close. A polite last note: "I will stop writing, but if timing changes, here is how to reach me." Then remove them from the sequence.
The AI task here is drafting four messages from the same profile with different hooks and checking they are not repeats. Ask it to list the single claim in each message and to flag any message that restates an earlier one. For the market-data side of touch two, source-checked reports and comps with ChatGPT shows how to keep comparables honest.
Before and after: removing the mail-merge feel
Before: "Dear Owner, I am a commercial broker specialising in your sector and would love to discuss opportunities for your property." After: "Hello [first name], I saw the roof works application for the building on the industrial estate. Have you had a chance to see what comparable units nearby are leasing at this year?" The second names a specific thing, asks one question and is shorter. The test is simple: could the message be sent to any other owner unchanged? If yes, it is not personal yet.
A second owner: an institutional landlord with a long hold
Not every target is a small private company. Say the next owner is a fund-backed entity holding a 60,000 sq ft office block bought 12 years ago, with the registry showing a change of directors this year and a refinancing charge filed 11 years ago, which suggests a debt term near its end. These are illustrative facts, but they show how the profile changes with owner type. For institutional owners the decision maker is rarely the name on the record; it is an asset manager, and the registered address is a service address. The profile's "next action" line becomes: identify the asset manager through the owner's website, press releases or a mutual contact, and do not guess an email address.
The message changes too. Smaller owners respond to a specific observation about a building; larger ones respond to a relevant comparable and a clear, short reason to talk. A draft might read: "Hello [first name], I track office lettings in the submarket and have three recent deals comparable to your block. If it helps your annual review, I can send the figures. No obligation." The prompt tells the tool to keep the message under 70 words, to avoid any claim about the owner's debt (you inferred it, you do not know it) and to offer something that stands on its own whether or not they reply. An inference like the old charge is useful for your timing and wrong to quote.
Checking the research before the message goes out
Before any first message, run a five-point check. Each point takes under a minute and prevents the errors that cost the most credibility.
- Is the owner still the owner? Re-open the record and check the date. A sale last month makes your message embarrassing.
- Is the entity correct? Similar names are common. Match the entity number or registered address, not just the name.
- Is the hook true and current? A planning application may have been withdrawn or approved; a unit "advertised for 14 months" may have been let last week.
- Is anything in the message an inference stated as fact? Reword it as a question.
- Is the contact appropriate? Check the suppression list and the rules for the channel.
You can ask the tool to run this list against your draft and notes: "For each sentence in the draft, say which note supports it, or write UNSUPPORTED." Treat UNSUPPORTED as a stop sign. The tool is checking your draft against your own notes, which is a task it does well; it is not checking the world.
A realistic mistake and how it showed up
A broker used a tool to research "owners of industrial buildings on a named street" and pasted the generated list straight into a mail merge. The list mixed genuine records with plausible names the tool had produced, and two letters went to companies that did not own anything on the street. The error showed up when a recipient replied asking how the broker had got their name. The fix was a rule: every name in a profile must carry a record citation in the notes, and the tool is only ever asked to organise notes the broker has gathered. AI does not look owners up for you. It works on what you bring.
Time and cost for a realistic week
Say a broker allots four hours a week to owner prospecting. Without a routine: two hours scrolling records, one hour writing messages from scratch, half an hour on logging and half an hour lost to switching, for about 10 messages sent, most of them generic. With the profile template and the drafting prompts: 90 minutes of research for six owners, 60 minutes drafting and checking six first messages and four follow-ups, 30 minutes on logging and the suppression list, leaving 60 minutes for calls. The output is ten messages again but each carries a verified, specific hook, and the follow-ups are drafted in advance. The cost of the AI part is a subscription at list price in the low tens of dollars a month for an individual plan; check the current price of whichever tool you choose. These are illustrative figures, and the real gain is the quality of the first line, not the volume.
Compliance limits to respect
This is practical guidance, not legal advice. Rules on unsolicited contact vary by country, state, channel and recipient type, and an owner who is a company may be treated differently from an individual. Before you send anything:
- Know the rules for each channel. Email, calls, texts and letters are often governed differently; text and automated calling are commonly the strictest. See the rules for calling and texting leads for the residential-side picture, and check what applies to you.
- Keep a suppression list. Anyone who asks to stop is removed from every sequence the same day, and the list is checked before each send.
- Identify yourself clearly. Your name, firm and a way to opt out in every message.
- Data protection. If you have contacts in the EU, data-protection law may apply to the profiles you keep on named individuals. Hold only what you need, note the source and delete when it is no longer needed.
- Do not misrepresent. No false urgency, no claims you cannot support, no implying you act for a buyer you do not have.
- Licensing and advertising rules. Check how your licence and your firm's policies treat prospecting messages and any required wording.
Add a line to your process: nothing is sent by AI. You read, edit and send every message, and the tool never has access to your mail account. That one rule keeps your name on what is actually sent.
What to put into the AI tool and what to keep out
Keep prompts to public facts, your own observations and templates. Leave out private information about individuals that you did not get from a public source, anything a client has told you in confidence, and any other party's financial details. Use a business plan that does not train on your content by default if you use confidential deal notes at all, and turn off the training setting on a personal plan. The comparison of AI tools for commercial brokers by job covers which tools suit research, drafting and CRM work.
Pace, tracking and how to tell it is working
Run the routine weekly: one hour to research five owners, 30 minutes to draft first messages and check them, and ten minutes to log replies and update the suppression list. Track three numbers per month: owners researched, first messages sent and replies of any kind. A realistic early pattern is a low reply rate on cold first touches and a higher one where the hook was specific and true; do not trust a published reply rate as a benchmark, because yours depends on your market, your list and your wording.
Review each quarter. Which hooks got replies? Which owners scored high but never answered? Adjust the scoring weights and keep the profile template. When an owner does reply, a deal-week routine for brokers picks up what to do next. If you would like help setting up the profile, the scoring sheet and the sequence for your own patch, that is what my AI implementation consultation covers, but the steps above work without it.
Questions brokers ask about AI owner prospecting
Can AI find who owns a property?
Not reliably by itself. Ownership comes from public records and your own data source, which you supply. AI is useful for organising what you found, spotting patterns such as an owner with several nearby assets, and drafting the outreach. Always confirm ownership in the record before writing to anyone.
Is it acceptable to email owners I have never spoken to?
Rules on unsolicited contact differ by place, channel and whether the owner is a company or an individual. Check the rules that apply to you for email, phone and text, keep a suppression list, and give a clear way to opt out. Ask a qualified adviser if unsure.
How many owners can I research in an hour?
A realistic pace is four to six owners an hour with a template, including the time to check the record. Faster usually means skipped checking. A short list of 20 well-researched owners beats 200 names with a mail merge.
What should never go into an AI tool?
Private information on individuals that you obtained outside public sources, financial details given to you in confidence by clients, and anything a client asked you to keep quiet. Keep prompts to public facts and your own notes.
Further reads
- AI Commercial Real Estate Underwriting: A First-Pass Model — A first-pass model for a property you pitch
- AI Automations for Real Estate Wholesalers: Sourcing to Contracts — Sourcing methods from the residential side
- AI Automation Agency for Real Estate: Scope, Fees and Red Flags — What an automation agency costs in this field
- Virtual Assistant or AI for a Real Estate Agent: What Goes Where — What goes to a person and what to AI
- Follow Up Boss Pricing: What a Small Team Really Pays — What a follow-up CRM costs a small team
- Can AI Abstract a Commercial Lease With Page Citations? — A lease abstract template, a prompt that forces page-cited quotes, a spot-check method and an exceptions list for the clauses AI most often misreads.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: public land-record and company-registry practice described in general terms; no vendor prices are quoted. Check the rules for unsolicited contact that apply to you.