
Photo by MART PRODUCTION on Pexels
AI in Real Estate 2026: 97,000 Conversations and What They Really Mean
This material was prepared with the help of artificial intelligence and checked by a person. Editorial responsibility: Aster Of Asia Co., Ltd..
Responsible for content: Leonid Ustinov, Aster Of Asia Co., Ltd.
Aster of Asia editorial team
Compass gave its AI assistant to 83,000 agents. A few weeks after the July launch, about 15,000 of them had used it, generating more than 97,000 conversations. The number sounds impressive until you do the second calculation: roughly 68,000 agents never touched the tool at all.
That is the real state of AI in property at the start of 2026. Not an algorithm replacing the agent, but a widening gap between those who have rebuilt their working day and those still logging clients by hand.
For Thailand the gap is even more visible. Most deals here run through small brokerage teams with no enterprise-grade CRM and no budget for a large rollout.
Quick Answer
- 83,000 agents got access to the Compass AI assistant. Around 15,000 used it in the first weeks after launch, which is about 18% adoption.
- Those 15,000 agents held more than 97,000 conversations, which points to repeat use rather than one-off experiments.
- The assistant combines about 100 industry tools in a single window. The core value is not text generation but removing the need to switch between tabs.
- The key shift is from chat to action: the system finds potential clients in the database, drafts personalized messages and transcribes voice notes after a viewing.
- What AI does not do: judging the price of a specific property, handling a buyer's emotions and taking responsibility for the deal all remain with the human. Compass says so openly.
- For the Thai market, the main 2026 gain is faster property checks and faster preparation of materials for remote international buyers, not automatic selling.
Key Facts
- Compass launched its AI assistant in July 2026 with a live public demo, then kept promoting it inside offices. Adoption took a dedicated campaign, not a mass email.
- The difference from the previous generation of AI tools is described as a move from content generation to agent-style actions: building client lists, drafting outreach, taking voice notes.
- The developers deliberately avoided branded interface jargon. Agents are taught to phrase requests in plain words instead of memorizing feature names.
- Compass leadership publicly acknowledges two risks: agent resistance, and a scenario in which buyers close deals directly through chatbots, bypassing the broker.
- About 82% of agents at the largest US brokerage network by headcount did not engage with the tool in the first weeks. That is an honest adoption benchmark.
- On the Real Brokerage side, the Leo 2.0 beta (running on the reZEN platform) targets automation of lead generation, marketing and back-office work for thousands of agents. The stated direction is automating roughly 80% of repetitive processes by 2027 while keeping agents for expertise and local knowledge.
- A Phuket agency case study (PhuketStayPro) shows an operation running almost entirely on AI tools, including a bilingual website and auto-posting to Threads, yet hundreds of automated posts did not equal conversions. Deal decisions stay with people.
- In Thailand, remote deals with foreign buyers have become routine, so video viewings and digital document packs are now the norm. This is the part of the workflow where AI saves the most time.
How to Start: Step by Step
- Measure where your week goes. Take 5 working days and log hours: viewings, messaging, preparing shortlists, document checks, owner reports. Many agents in Phuket and Bangkok spend 12 to 18 hours a week on messaging and shortlists. Automate that block, not the viewings.
- Start with one task, not a platform. Compass invested in a live demo and office campaigns and still reached only 18% adoption in the first weeks. A five-person team will not manage a big-bang rollout. Pick one process, such as the follow-up email after a viewing, and make it automatic within a week.
- Build your own knowledge base. Put your project terms in one document: price per square meter, handover dates, sinking fund, common area fee, foreign quota status for each building. Without it, any AI will confidently invent answers. With it, you can answer a client in 30 seconds instead of 30 minutes.
- Set up voice transcription. After each viewing, dictate three minutes: what the client liked, what worried them, budget, timeline. The system turns it into a structured note. It is the most underrated feature in the Compass set and the only one that pays off from day one.
- Run document checks as a checklist, not a chat. Chanote title deed, foreign quota confirmation, utility arrears statement, ownership history. AI is good at comparing a document against a checklist and spotting mismatches in names and areas. It does not replace a lawyer and carries no responsibility for the result.
- Prepare the client's viewing tour end to end. A 3-day route, 6 to 8 properties, logistics between Laguna and Rawai accounting for traffic. AI drafts the schedule in a minute. Book flights and hotels early, since rooms near viewing areas almost sell out in high season from December to February.
- Measure one metric after a month. Not the number of AI conversations, but the time from the client's first inquiry to signing the reservation agreement. If it has not shortened, you automated the wrong step.
FAQ
Will AI replace real estate agents in Thailand?
Not in the coming years. Compass, where thousands of agents use the assistant, states plainly that judgment, emotional intelligence and client relationships are not reproduced by a machine. Thailand adds its own specifics: title verification, negotiating a discount with the developer, and knowing which building will actually be delivered on time. These are not tasks for a language model.
We will shortlist properties for your budget
Pick a range and we will send a shortlist with prices, layouts and payment plans within 24 hours.
What does AI already do better than a human?
High-volume routine work: sorting client databases, drafting emails, transcribing notes and a first-pass comparison of documents against a checklist. Compass packed 100 tools into one window precisely to save on switching.
Why did 82% of agents not adopt the new tool?
Because adoption is limited by habit, not technology. An agent with eight years of experience already has a working process and will not change it for a promised saving. What breaks that inertia is not a mass email but a colleague at the next desk getting a visible result.
Should a foreign buyer close a deal directly through a chatbot?
The scenario is openly discussed and the risk to brokers is real. But in Thailand the cost of an error is high: foreign quota verification, the developer payment structure, and currency transfers with the correct payment reference for later registration. A bot will not inspect the construction site or notice that the neighboring building has been in court for two years.
How much does it cost to start using AI as an agent?
A basic setup runs about 20 to 60 US dollars per month per person. The expensive part is not the subscription but the 10 to 15 hours spent building a knowledge base for your own projects. Without that work, the tool produces generic phrasing that clients recognize instantly.
Can I trust AI to check Thai property documents?
Only as a first filter. The model finds discrepancies in areas, names and dates. A legal opinion on the Chanote and quota status comes from a Thai lawyer, and the lawyer signs it, not the algorithm.
Where should an investor, not an agent, begin?
By checking the other side. Ask the agent to show how returns are calculated: occupancy by month, management fee, furnishing and replacement costs, tax. If the answer is a single figure like 7% a year with no breakdown, there is no calculation, manual or machine.
The takeaway after a year of watching these tools is simple: the winner is not the one with the most expensive subscription but the one who first digitizes their own knowledge of specific buildings, prices and deadlines. If you sell fewer than three properties a year, do not spend time on this. The manual process will be faster.
Source: Fortune
Ready to invest in Thailand? Our experts will help you find the perfect property.
Want to master AI tools for real estate? We offer a free course with practical AI skills for property professionals: Enroll for free - https://class.asterofasia.com/
Ready to start?
Answer 4 questions and we will prepare a personalised selection of property in Thailand.
What is your goal?