How estate agents in the UK are using AI in 2026

Walking onto the floor at any UK estate or lettings agency in 2026, you'll see AI showing up in three places. Drafting property descriptions. Triaging inbound enquiries. Generating viewing summaries for landlords. The tools have settled into a recognisable shape and the early novelty has faded, which is a good thing. It means there are now patterns to copy rather than experiments to debate.

This is a field-level look at what's actually working in UK agencies right now, what isn't, and where the time savings land.

The three jobs AI does well

Property descriptions

ChatGPT, Claude, and the property-specific layer in Reapit's recent releases can take a Rightmove-style spec and turn out a usable 200-word description in seconds. The catch is style. Most agents end up writing a one-paragraph house style guide and pasting it into the prompt every time, because the default tone is too American and too breathless for a UK audience. Set the rules once and the time saving is real.

The agencies getting the most value have built a small library of saved prompts: one for sales, one for lettings, one for HMO, one for new-build. Five minutes of setup, then it's a paste job for every listing. We've seen senior negotiators get an afternoon back per week from this alone.

The mistake to avoid: trusting the first draft. Every AI description needs a human eye on it before it goes live. Wrong number of bedrooms, fictional features, or a description that contradicts the photos will damage trust in ways the time saving doesn't make up for.

Enquiry triage

Front-of-funnel chatbots from Chatbase, DESCRIBLY, and a handful of property-specific vendors can answer the standard "is the property still available", "what's the deposit", "can I bring a dog", "when can I view" questions without a human in the loop. The good ones are trained on your listing data and your tenancy criteria and stop the conversation cleanly when they hit something they can't answer, handing off to a real person.

The bad ones hallucinate, contradict your terms, and embarrass you publicly on a property listing or a Facebook ad reply. The difference is almost entirely about how carefully you set up the knowledge base. Two days of work scoping the bot's boundaries up front saves a year of cleaning up after it.

The metric agencies actually care about is qualified viewings booked per pound spent. A well-trained chatbot can handle 60 to 70 per cent of inbound traffic without a human, which means your team only ever speaks to people who are properly interested.

Viewing summaries

Recording the call or the in-person viewing and getting an AI-generated summary back, with action items, has quietly become the highest-leverage workflow in lettings. Landlords get a same-day report, the agent doesn't have to type it, and the file is searchable. Tools like Otter, Fireflies, and the Microsoft Teams summary feature all do a reasonable job. Property-specific layers from Goodlord and Reapit are catching up.

The agencies running this well share two habits. They standardise the format, so every summary has the same sections (condition notes, questions raised, next actions), and they file the summary into the property's record automatically rather than leaving it in someone's inbox.

Where AI still falls short

Negotiation. Compliance checks. Anything where the consequence of being slightly wrong is large. The pattern across the agencies we've spoken to is clear: AI handles the volume tasks, humans handle the judgment tasks, and the tools that try to do both end up trusted for neither.

A lettings agent in the South West summed it up: "I'd let AI write a hundred descriptions before I'd let it write one offer letter." That's the right posture.

The other failure mode is over-fitting the workflow to the tool. We've seen agencies redesign their property onboarding around a specific AI feature, only for the vendor to change pricing or get acquired six months later. Build for the job, not the brand.

What this means for your agency

If you're starting now, the order to adopt is descriptions first, then summaries, then triage. Descriptions show value in a week. Summaries take two weeks to embed but become the thing nobody wants to give up. Triage is the most powerful and the most likely to go wrong without proper setup, so leave it until you've got the first two working.

The total time investment for an agency to get all three running properly is around two weeks of focused effort. The annual time saving for a five-person team usually clears 800 hours.

If you want a personalised starting plan for your office, the AI Opportunity Assessment is a £19 pilot assessment delivered within 24 hours. It shows which opportunities are worth exploring first and which ones to leave alone for now.