Welcome! Let's start with an honest question
Have you tried an AI assistant at work yet? Maybe ChatGPT, Gemini, Claude or Microsoft 365 Copilot? If you have, you've probably felt two things at once. First, wow, that was quick. Then, a moment later, hang on, is that actually right?
Both of those feelings are useful. In this level we'll work out where AI really helps you in agency and lettings work, and where it lets you down. We'll also look at what recent surveys say about how agents are using it. And you'll learn to read those surveys with a careful eye.
What is an AI assistant, really?
Here's the simple version. Tools like ChatGPT, Gemini, Claude and Copilot are built on large language models. They learned from huge amounts of text, and they write by predicting which words are likely to come next.
Think of a very well-read new negotiator on their first day. They've read thousands of property adverts and emails, so they can write in that style. But they've never been inside the house you're selling. They don't know the boiler is new or that the garden faces north. Unless you tell them, they're guessing.
That's the key idea for this whole course. AI is great at language. It doesn't automatically know your facts.
Where AI genuinely saves you time
So where does that well-read new starter shine? On work where you supply the facts and it handles the wording.
The first is first drafts of property descriptions. You give it your notes from the valuation, the room list and the features, and it turns them into flowing copy you then edit.
The second is routine emails and messages. Think of a polite reply to a viewing request, or a chaser to a landlord about a gas safety certificate. It's quick to draft and quick for you to check.
The third is admin and social media. It can tidy up meeting notes, turn a checklist into a clear email or suggest three caption ideas for a new listing.
What do these have in common? You know the truth, and you can spot a mistake in seconds.
Where it can't be trusted on its own
Now the flip side. Some tasks need facts the tool doesn't have, or judgement it can't give.
One is facts about a specific property. Tenure, council tax band, room sizes, lease length: an AI tool may fill in gaps with something that just sounds right. A made-up detail in an advert is a real problem, not just a typo.
Another is valuation figures and comparables. Ask a general chatbot what similar houses sold for on a street and it may give you neat, specific numbers that don't exist. Comparables have to come from real sales data that you check yourself.
The last is legal and compliance answers. Rules for agents and landlords change, and a tool may give you an out-of-date or invented answer in a very sure tone. Your firm's policy and official guidance come first. We'll dig into those rules in later levels.
Why does AI sound so confident when it's wrong?
Good question, and here's the plain answer.
The model is built to produce fluent, likely-sounding text. Fluent isn't the same as true. When it doesn't know something, it often doesn't say "I don't know". It produces the most plausible-sounding answer instead. People call this a hallucination.
OpenAI's own researchers have explained that standard training and testing tend to reward guessing over admitting uncertainty. It's a bit like a multiple-choice exam where a blank scores zero but a guess might score a point, so guessing pays. They also say newer models hallucinate less but still do it.
So a confident tone tells you nothing about accuracy. You're the one who checks.
What does the evidence say about agents using AI?
Let's look at some real numbers, with dates.
In October 2026, the property software firm Alto published a survey of 250 UK estate and letting agents. It found that 69% use AI regularly or every day. The most common use was writing property descriptions (67%), then emails (53%), admin (37%) and marketing and social media (35%). Far fewer used it for valuation preparation (10%) or compliance (5%). And 64% said they worried about AI making mistakes.
Interesting, right? It matches the split we just walked through. Agents lean on AI for words, not for facts and judgement.
How to read a statistic before you repeat it
Imagine a colleague says in a meeting, "82% of agents use AI!" Should you quote that to a vendor?
First, check three things: date, source and sample.
That 82% figure does exist. It comes from a survey of 225 US real estate professionals, published in early 2026 by Realtor Property Resource, a company linked to the US National Association of Realtors. So it's American, it's a small sample and it came from a company inside the industry. It's not evidence about UK agents.
The Alto survey needs the same treatment. It was published in October 2026 by a company that sells software to agents, and it asked 250 people. That's useful, but it's not the final word. Even news reports of the same survey don't always agree. One said 76% worried about losing the personal touch, and another said 75%. When it matters, go back to the original source.
Your turn
This week, pick one AI-related statistic you've seen in a trade article or on LinkedIn. Write down its date, who published it, how many people were asked and in which country. Then decide whether you'd repeat it to a client.
Your checklist:
- Is the AI doing the wording while I supply the facts?
- Have I checked every property fact, figure and legal point myself?
- Do I know the statistic's date, source and sample?
Key takeaways
- AI is strong at drafting descriptions, emails, admin and social posts from facts you give it.
- On its own, don't trust it for property facts, comparables and valuation figures, or legal and compliance answers.
- AI can sound confident and still be wrong because it's built to produce likely-sounding text, and guessing often gets rewarded over saying "I don't know".
- An October 2026 Alto survey of 250 UK agents found 69% use AI regularly, mostly for writing tasks.
- Before you repeat any statistic, check its date, source and sample.