AI in Dubai real estate: what changes for brokers, investors and developers

Which jobs does he expect to disappear first?

Anything that involves tapping away at a keyboard, within two to three years. Artashes expects a large redistribution in the labour market, with fewer people able to do more work and outdated skills left behind.

Skilled manual craft is safer in his view. His example is a Japanese knife maker using a technique passed down through generations, which is very hard to teach to a machine.

Driving is another matter. He pointed to driverless cars already operating in San Francisco, and both speakers mentioned autonomous trucks in the United States and China.

Voice instead of keyboard: the first efficiency gain

The first thing he changes is typing. There are apps where you speak into a microphone and the text appears wherever the cursor is. He estimates that simply giving up the keyboard makes a person about 80% more efficient.

His explanation is psychological. When typing, people bargain with themselves about whether to write more or stay brief. When speaking there is no bargaining, so far more detail reaches the AI for the same effort.

Speech also skips the internal filter. He compares it to a psychologist's test where you answer as fast as possible. Oleg noted that some writers dictate their books for the same reason and edit afterwards.

What can a broker build with AI in two days?

A working personal website on day one and a project of their own on day two. In his two-day intensive, participants first dictate a full A4 page describing the site they want, then generate it and see the buttons, navigation and animation actually work.

Many struggle to dictate more than six lines, so he pushes them to a full page. Along the way they meet the terminal, the black screen with text that people assume is only for IT specialists, and get past that fear.

The most striking personal project he named came from a 62-year-old participant. He built a content pipeline that produces short videos by itself, using an AI avatar and a cloned voice connected through an API.

A custom CRM built from WhatsApp chats is possible in theory but hard in practice. Cold messaging without being blocked requires passing verification on a Meta business account, a long process that people do not finish during the course.

Which AI automation pays for itself with a single deal?

A broadcast to your own contact list. Artashes named it without hesitation. Re-engaging old clients by hand is slow, and a copied message reads as dry. AI can analyse each conversation and write a personalised message for every client.

Oleg called the manual version titanic work. The guest said building such a tool takes about a day and adds a lot of efficiency, because nobody has to reread what each client wrote before.

Why does AI still get property facts wrong?

Because it invents details and uses stale data, and because clients do not know how to ask. Artashes says an AI that is asked a poorly framed question tells the person what they want to hear.

Oleg gave two examples from practice. One client insisted, on ChatGPT's word, that a construction site in Al Satwa used to be a cemetery. They searched old maps until they proved there had never been one.

The second is prices per square foot. Whether the question is about Thailand or Dubai, the assistant may return a figure from years ago that is completely off the current market.

He also described a very large deal that collapsed. The buyer's lawyers consulted AI and demanded changes to standard, non-negotiable clauses of the developer's sale agreement, plus a shorter penalty-free delay period, for a building that was essentially already standing.

Both speakers agreed that property law is tied to the country where the deal happens, with its own courts, practice and cases. Generic advice from a distance, human or artificial, tends to break deals more than protect them.

How should an investor use AI when buying property in Dubai?

First to choose one broker, then to check what that broker says. Artashes suggests putting ten candidates, with names, phone numbers and emails, into a deep research tool and asking which one it would work with, favouring people with a visible track record.

He argues for a single broker because ten brokers compete against each other. When one offers something, the other nine try to kill the option, and the buyer ends up unable to choose anything for three months.

After that, everything the chosen broker says also goes into deep research before a decision. Oleg's example investor was a small owner with three apartments worth about USD 1 million each, not a fund.

Two skills decide the quality of the answer: giving the AI accurate data, and asking a good question about that data. Artashes says those two skills solve the buyer's problem.

Oleg sees potential in services that scan listing portals in real time and show how much cheaper a unit is than the next option. For an investment studio or one-bedroom, entry price and return matter more than floor or view.

The limit is qualitative. Two units can have a similar price per square foot and original price, yet one faces a courtyard and the other has a view, with a 30% price difference. AI handles numbers brilliantly; view and condition still need a person.

How AI changes developers and their sales teams

Artashes expects little change in pricing itself: a market is a market, with supply and demand competing as before. What changes is cost. How far developers integrate AI, and how much their costs fall, will shape competition between them.

He sees AI as already very effective in design and architecture, and is less sure how it applies to construction itself.

Oleg expects sales teams to shrink. A website chat would give the first consultation on location, price and handover date and, in his estimate, filter out 98% of enquiries before real people hold the meetings.

The guest's key point is infrastructure. Large companies already store listings, floor plans and a CRM. Developers who add a text interface that lets a buyer's AI agent query that data and find the best apartment will win.

He compared it to the race to be named by ChatGPT as the best broker in a given community. People will search less and ask an assistant more, including by voice from a car, so being the answer becomes a new advertising channel.

What happens to large real estate agencies?

They will have to adapt, and nobody can yet picture how. Artashes expects fewer brokers and less reason for a strong broker to sit inside a large company, but he also expects companies to build services for brokers who will not build their own.

Oleg argued that AI reduces the need for a back office and that an agent's personal brand now matters more than the company's. The guest replied that he has not met anyone whose transaction paperwork truly runs on AI.

The obstacle is documentation. A conveyancer knows the steps in his head, but to hand the job to AI every step must be described precisely. Those protocols do not exist, and they cannot be written for one small link in isolation.

They have to be designed from the top of the company, or the pieces will not connect later. So the change must first happen in the head of the person in charge.

Oleg sees a niche for firms built around proprietary software that scans portals and answers clients, where a broker gives up half the commission but closes three deals a month instead of one. The guest's caution: any software lead may last two weeks.

Deepfakes and fraud: the other side of the technology

Oleg raised the risk of scammers using AI, from a fake video avatar approving a transaction remotely before a notary to forged documents. Artashes answered that crime has always moved forward hand in hand with progress.

He expects the next step to be detection tools, for example a messenger that labels generated content on screen. Both compared the moment to earlier technologies that were first met with bans and odd rules before the market settled.

Where should a broker start with AI this evening?

By teaching the assistant who you are. Open a paid subscription to ChatGPT or Claude and tell it to ask you 30 questions, one at a time. After that it talks to you as you, not as a random person.

The first agent to build is a personal assistant. Oleg noted that almost all top brokers already have a human assistant, so an AI one is a natural first step.

Artashes warned against starting big. He once had AI create a department of 47 agents covering health, nutrition, relationships, child care and more, and ended up not understanding how any of it connected.

His lesson is to build small pieces and get used to each one before combining them. A large system created by something else is one you cannot operate, however well it was assembled.

He also argued that the value has shifted from stored knowledge to the will to act. With AI all the world's information is at hand, and the skill is learning to narrow it down. In his words, the most important tool is still your own head.

What happens to those who ignore AI?

They fade out the way road atlases did. Artashes recalled that when smartphones brought navigation, every car still had an atlas in the glove box, and for about ten more years its users either switched or stopped driving.

Oleg added the taxi example. For a while some people booked through an app and some still phoned a dispatcher, and then the phone option disappeared almost overnight.

Experienced brokers with twenty years of personal clients and several investment cycles behind them are in less of a hurry. The guest thinks that for them it may not be a risk, since many own rental apartments. Oleg expects younger, hungrier agents to push them out anyway.

Who wins, according to Artashes: the curious, those with an inner drive to figure things out. The world is changing so fast that people who cannot change with it will lose.