Practical guide
AI for the real-estate agent: more sellers, more buyers
What you do, what the AI does, and how you know it is working. No promises, just measurement.
AI does not bring you clients. It gives you back the time for them.
Every deal goes through the same steps. AI speeds up the preparation for each one. The conversations and the viewings stay yours.
- 1conversation
- 2viewing or meeting
- 3contract
- 4deal
Everything in this guide serves one number: how many real conversations you have with people who genuinely want to sell or buy.
What you need
- ChatGPT or another AI assistant on your phone and laptop. The free tier is enough to start.
- 20 minutes a day at the same time every day, for example with the first coffee.
- Your real data: past clients, enquiries from your listings, the properties you represent.
Nothing new to install, no subscriptions on day one. You start with what you already have.
Who does what
| The AI does | You do |
|---|---|
| Drafts messages, listings and posts | Read, edit, press send |
| Sorts enquiries and buyer questions | Hold the conversation and the viewing |
| Collects and compares property information | Check the numbers before you say them |
| Reminds you whom to call today | Pick up the phone |
| Writes fast and a lot | Decide what is true and what is right |
The golden rule: not one AI sentence reaches a client before you have read it.
The map: where clients come from
Sellers
- Past clients and the people they know
- Owners who are selling on their own
- People who follow you for useful local posts
Buyers
- Enquiries from your listings on Facebook and the classifieds portals (in Bulgaria: OLX, alo.bg, bazar.bg)
- Referrals from happy buyers
- Your list of active searchers
Five concrete paths follow: three for sellers, two for buyers. Each has an exact task for the AI and an exact task for you.
Past clients and referrals
The people you have already helped are your warmest source. The problem is never whom to write to. It is that the writing keeps getting postponed.
- Make a list: everyone you sold or found a property for in the last three or four years.
- The AI writes a short personal message for each one. You edit it until it sounds like you.
- Three messages a week. No more. Personal, never bulk.
Measure: messages sent, replies, conversations about a property.
A personal message to a past client
Help me write to a past client. I sold them [a two-bedroom flat in neighbourhood X] in [year]; their name is [name]. Write a short message for Messenger (under 60 words): I ask how they are, mention that the market in their neighbourhood has picked up, and say that if they or someone they know is thinking about selling, I would gladly help with a free valuation. Warm and natural, no sales tone, no exclamation marks. Give me two versions.
Before you send: Are the name, the property and the year correct? Does it sound like you? If not, tell the AI what to change, or write it in your own words.
Owners selling on their own
Every week owners list their homes themselves on the classifieds portals. Many of them look for an agent after months without a result. The agent who wins is not the one who writes to everybody, but the one who shows up prepared.
- Once a week, review the new owner listings in the neighbourhoods you work.
- Pick one or two. The AI prepares a call brief from the listing and your own comparable sales.
- Call in person, with specifics about their property, never with a template.
Rule: no bulk messages through the portals. If the property is already with another agent, that is respected.
A call brief for an owner
Here is a listing from an owner selling on their own: [paste the listing text]. Here are two or three similar properties I know: [price, size, how long they took to sell]. Prepare a short brief for a phone call: 1) what we know from the listing: price, size, floor, how long it has been online; 2) how it compares with my comparables; 3) three questions to ask the owner. Do not invent data. If something is missing, write "no data".
Before you send: The comparable figures are yours, not the AI’s. Check them. The brief supports the conversation; it is not a script to read aloud.
The neighbourhood post
An owner decides to sell months before calling an agent. In that time they watch who understands their neighbourhood. One useful post a week makes you that person.
- Collect three or four facts from your own work: what sold, at what price, how fast.
- The AI turns the facts into a short post. You remove anything you would not say out loud.
- Publish on the same day every week. Consistency is the whole trick.
Measure: not likes, but private messages and questions like "what is mine worth?".
A post about the neighbourhood
Write a short Facebook post (under 120 words) about the property market in [neighbourhood, city] this month. Here is my data: [for example: a three-bedroom sold for X €, average Y weeks from first viewing to sale, two-bedroom flats under Z € are in demand]. Tone: a helpful local expert, not an advert. No clichés like "dream home" or "unique opportunity". End with: "If you are wondering what your property is worth, ask. The valuation is free."
Before you send: Every number comes from your own work or a verified source. One wrong figure costs more trust than ten posts earn.
From enquiry to viewing
Most enquiries die because the reply comes late or leads nowhere. A fast reply with the right questions and a proposed time turns a curious person into a booked viewing.
- New enquiry: hand it to the AI together with the property details.
- The AI drafts a reply: thanks, an answer to the question, three qualifying questions, two proposed viewing times.
- You edit and send. The target is within two hours, while the interest is alive.
Measure: enquiries, replies within two hours, viewings booked.
A reply to an enquiry
I received an enquiry about my listing: [three-bedroom flat, neighbourhood, 145 000 €]. Here it is: "[paste the enquiry]". Write a reply: a short thank-you, an answer to their question, then three questions: what budget and financing they have (mortgage or cash), in what timeframe they want to buy, and what is non-negotiable for them. End with two proposed viewing times: [Wednesday 18:00] and [Saturday 11:00]. Polite and businesslike, no empty phrases.
Before you send: Are the price and the details correct? Do the times really suit you? Remove any question the person already answered in the enquiry.
The buyer list
Every buyer who does not buy today is tomorrow’s deal, if you remember what they are looking for. Paper and memory lose things; a simple table does not.
- After each conversation the AI summarises what the person wants: budget, neighbourhoods, must-haves. You add it to your table.
- A new property comes in: ask the AI "who on the list fits, and why?"
- A match means a personal call: "I thought of you, exactly what you are looking for has come up."
That call is the most pleasant one in the profession, and the most common reason for a referral after the deal.
The next level: agents
An ordinary chat answers one question. An agent carries out a whole task: it reviews, collects, sorts and brings you a finished result to check.
An example task for an agent
"Every Monday: review the new owner listings in [neighbourhood A] and [neighbourhood B], put them in a table with price, size, date and link, and flag the ones that have been online for more than a month."
- Start with one repeating task you already do by hand.
- Several agents only make sense for independent tasks: one collects listings, another drafts messages.
- Check an agent’s output the way you would check a new intern’s: useful, but not on trust.
The five rules
- Nothing goes out unread. You are responsible for every word. The AI only writes drafts.
- Check numbers, names and facts. AI mistakes look the most convincing.
- Few and personal beats many and bulk. Five prepared conversations are worth more than fifty copied messages. No spam through the portals.
- Other agents’ contracts are respected. A property listed with another agent is not a free property.
- Work with data people gave you themselves. And no invented claims in listings.
The week with AI
Every day · 20 min
- Reply to new enquiries: AI draft, your edit
- Follow up yesterday’s unanswered ones
Once a week · 60 min
- One neighbourhood post
- Three personal messages to past clients
- Review the owner listings, prepare one or two conversations
Under three hours a week in total. It does not take an athlete’s discipline, just the same slot in the calendar.
The experiment: two weeks, one thing
Do not start everything at once. Pick one path. The easiest start is "from enquiry to viewing". Measure it for two weeks.
- Week 1: only measure, working as you do now: how many enquiries, how many answered within two hours, how many viewings.
- Week 2: every enquiry goes through an AI draft first. The same three numbers.
- Compare. If the numbers grow, keep it and add the next path. If they do not, change the approach instead of doubling the effort.
- ?enquiries
- ?answered within 2 h
- ?viewings
- ?contracts
Nobody can promise what the numbers will be. That is why they are measured, not believed.
And finally
Kove
Everything above works with any AI assistant. Once your routine has settled, part of it, the enquiries, the reminders, the listings, can run through Kove: an assistant set up for the Bulgarian language and for an agent’s work.
No promises. Try it and judge for yourself.
Try Kove