AI Tools for Estate Agents: A Stack by Job, Valuation to Completion

Use a general assistant for valuation prep and listing drafts, an AI call-answering or booking tool for enquiries and viewings, a spreadsheet plus assistant for offers, and a progression tracker with drafted chasers through to completion. A small agency can start for roughly $20 to $50 a month and add paid tools only where a stage stalls.

Each stage has one thing the AI must not do. A valuation is an opinion the agent signs, a listing states only facts someone has checked, an offer is passed on exactly as it was made, and a progression message never promises a date the solicitors have not given. The stack below is built around those four limits, with list prices and a worked example at each stage for a four-person agency running about 25 sales in progress.

Follow me on Instagram@sagnikteaches

Valuation: comparables in a sheet, the brief in a draft

The job is to arrive at the appointment knowing the likely range and the vendor's situation. Use the assistant for two things: writing the spreadsheet formulas and drafting a one-page brief. Do not use it as the calculator.

Connect on LinkedInSagnik Bhattacharya

Export ten to fifteen recent sales of similar homes nearby from your CRM or from sold-price data you are licensed to use, with price, floor area and sale date. Ask the assistant to write formulas for price per square foot, the median and the quartiles, then run them in the sheet. For the example, ten comparables give these prices per square foot, sorted: 389, 398, 402, 405, 412, 415, 420, 425, 431 and 440.

Subscribe on YouTube@sagnikteaches

That gives a working range of about $442,000 to $468,000 before anything about the particular house has been considered. The adjustments are the agent's: the kitchen's condition, the garden, parking, aspect, noise, and what sold fastest. The assistant then drafts the appointment brief from the sheet and the property facts: a comparables summary, three questions for the vendor (why they are moving, their timescale, whether they are in a chain), and the fee options. The tutorial on preparing valuation appointments with AI covers the brief in more depth.

Three watch-outs matter here. Check that every comparable measures floor area the same way, since one internal measurement among gross ones skews the median. Tell the assistant to use only the supplied sales, because otherwise it may add "market trends" that are not in your data. And keep the sheet in the instruction file as the working behind the figure you give, so the number you quote can always be explained.

Listing: particulars, photos and a facts check before it goes live

The job is a description, a feature list, a floor plan and photos that a buyer can trust. Three kinds of tool help. CRM-native drafting is one: Street's page describes an AI onboarding tool that reads uploaded photos and floor plans and drafts a listing, and Reapit's page describes AI-generated property descriptions with a setting for your own tone of voice. A general assistant working from a facts sheet is another. Photo tools for editing and virtual staging are the third.

Whichever you use, the control is a one-page facts sheet for each instruction, filled in by a person who has been to the property: tenure, floor area, number and size of rooms, heating and when it was last replaced, parking, outside space, services, known issues, and anything a buyer must be told. The assistant may use only that sheet, and it writes "[CHECK]" wherever a sentence needs a fact the sheet lacks.

Here is a before and after for a three-bedroom house, with an invented facts sheet:

Every change moves a vague claim to a fact on the sheet, and "ample" became "one car". Misdescription is the live risk with AI particulars, because a model will supply plausible features that the property does not have. Required property information is a separate matter: your market's rules say what a listing must disclose, and those facts come from documents, never from a model's guess. The checklist in material information in listings shows what to capture.

For photos, virtual staging is the paid extra most agencies consider. Virtual Staging AI's pricing page lists, on yearly billing, Basic at $16 a month for 6 photos ($2.67 each), Standard at $19 for 20 ($0.95), Professional at $39 for 60 ($0.65) and Enterprise at $79 for 150 ($0.53), with no watermark on any plan. The page advertises up to 50% off for paying yearly, so monthly billing costs more; check the toggle before you compare. If eight new instructions a month each get four staged photos, that is 32 photos, so Professional is the plan that fits: $468 a year, or $39 a month, about $1.22 for each photo you use. The page says nothing about disclosure, so write your own rule: every staged or digitally edited image is labelled on the listing, and the original sits in the file. The tutorial on AI photo edits without misleading buyers sets out how to word the label.

Enquiries and viewings: answering, booking and feedback

This stage is where an agency loses buyers by answering late. The tools split into call handling, booking, and what you do after the viewing.

For calls, Street's page lists an AI call handler that "answers your calls when your team can't". A standalone option is Smith.ai's AI receptionist, which is free for 25 calls a month and $3 a call after that, with paid tiers of $150 for 75 calls, $270 for 150 and $500 for 300. The break-even is worth knowing. At 60 calls in a month the free plan comes to 35 extra calls at $3, which is $105, against $150 for the 75-call tier, so you only move up once you pass about 75 calls. Above 75 calls the paid tier charges $2.50 for each extra call, so re-run the sum each quarter with your real call count. The wider call-flow questions are in the tutorial on an AI receptionist for estate agents.

Whatever answers the phone needs written rules. It books viewings, takes a name, number, property and preferred slot, and confirms by text. It states price or availability only from a live CRM field. And it hands any offer, valuation question, complaint or "is it still available?" about a property that has just gone under offer to a person immediately. Test it first with exactly those calls.

After a viewing, send the buyer three questions by text: what they liked, what put them off, and whether they would come back. The assistant is useful for the weekly vendor update, but only when the counts come from a sheet. Suppose a week brings 14 viewings and 9 replies: five praise the garden, four the kitchen, three say the asking price felt high and two mention road noise. The assistant drafts the update from those counts and from three verbatim comments you pick. The rule is that criticism is reported at the same weight it was given, and a vendor update that turns "price felt high" into "buyers are engaged" is the failure to prevent. For debrief calls with a negotiator, note-takers such as Otter Pro ($8.33 a user a month billed annually, $16.99 monthly) or Fireflies Pro ($10 annual, $18 monthly) save typing, though you should ask your compliance lead about recording consent. Rules for vendor updates and feedback go deeper on the wording.

Offers: log them exactly, summarise them, never reword them

The job is to record every offer as it was made, pass it to the vendor promptly, and keep a clear comparison. In some markets passing offers on promptly and truthfully is a legal duty, so ask your compliance lead what applies. The assistant's role is narrow: it summarises an offer record that a person has already written down, and it never negotiates, ranks or recommends.

The record comes first. The negotiator types each offer into the CRM within the hour, copying figures and conditions from the buyer's own message and noting the time received. The prompt then says: copy every figure exactly, no "around" or "roughly", list the conditions the buyer stated, do not rank, do not recommend. Take a guide price of $450,000 and three offers, all invented:

From that table the assistant drafts the vendor email, and the negotiator checks each figure against the three original messages, which takes under a minute per offer:

Three offers have come in on [property address] since Monday, all in writing and attached. A is $445,000 from a cash buyer with no chain, looking to exchange in about four weeks. B is $452,000 from a buyer whose own sale is under offer, with a mortgage agreed in principle and about twelve weeks to exchange. C is $448,000 from a first-time buyer with a mortgage agreed in principle and no chain, with about ten weeks to exchange. Let us know which you would like to accept or negotiate, and we will talk through the risks of each.

Do the small sums in the sheet, not the chat. B is $7,000 above A, which is 1.6% of A's price, and $4,000 above C. Whether that premium is worth a chain and an extra eight weeks is the negotiator's advice and the vendor's decision, and the draft takes no position. Keep identity and funds evidence for each buyer in the same file. In many markets estate agents have anti-money-laundering duties, and the assistant can help name and sort documents, but it never decides whether a check is satisfied.

Progression: a tracker, a stall flag and drafted chasers

From sale agreed to completion, the work is keeping every file moving, and the daily grind is chasing. Two tool types fit. CRM progression modules are one: Reapit's sales CRM page mentions sales progression, and Street's page lists an AI sale summary that summarises a sale's emails and notes and suggests next steps. The other is a spreadsheet tracker with an assistant to summarise threads and draft chasers.

The tracker carries one row per sale with these fields: property, buyer's solicitor, memorandum of sale sent, identity checks complete, searches ordered, mortgage valuation booked, survey, enquiries raised and answered, contract issued, contract signed, target exchange, target completion, date of last contact, next action and owner. The stall rule is the useful part. A file is flagged when nothing has moved for ten days or any milestone has passed its target date, and the flag tells the progressor which files need a phone call rather than another email.

Now the numbers. With 25 files and three chasers per file each week, writing every chaser by hand costs about six minutes: finding the thread, remembering where the file stands, wording it. That is 450 minutes, or 7.5 hours. If the assistant drafts from the tracker row and the latest thread and the progressor reviews and sends in about two minutes, the same chasing takes 150 minutes. The saving is five hours a week, or about $125 at $25 an hour. Those are assumptions, so time your own team for a fortnight first.

A drafted chaser should state facts and ask one question:

Hi [first name], following up on [property address]. The buyer's mortgage valuation was booked for [date] and we have not yet seen confirmation it has taken place. Could you tell us whether the report has arrived and what is outstanding before you can issue the contract? Our vendor is aiming to exchange by [date]. Thanks, [your name]

The rules for chasers are short. Dates are copied from the tracker and never worked out by the model. A message never promises or predicts an exchange or completion date, and never tells any party what a legal document means. After a third unanswered chaser, a person phones. When you use a thread summary, check who said what and when, because summaries can merge two senders. The step-by-step version of this stage is in sales progression chain chasers with AI.

Completion day: the last messages and the review request

The assistant can draft the completion-day messages, but a person sends them only after the conveyancers confirm that funds have arrived and keys can be released. A draft that says "completed" before that confirmation is the one error here with real consequences, so make the tracker's completion field the only trigger. After completion, a short thank-you with a link to leave a review is fine to draft. Send it only to genuine customers, offer nothing in return for a review, and never write a review on a customer's behalf, because platform rules differ and a faked review damages the agency's name. Keep the whole file, including offers, messages and identity evidence, for as long as your retention policy says.

What the whole stack costs for a four-person agency

The shared assistant is the one cost that scales with headcount. ChatGPT Business and Claude Team are each $20 a user a month billed annually or $25 monthly, with a two-seat minimum, and Microsoft 365 Copilot Business is $21 annually or $25.20 monthly. Business plans of this kind don't train on your content by default, which matters because the facts sheets, offer records and chasers contain clients' details. Here is the sample agency's stack, using the prices quoted above:

That adds up to about $201 a month with everything billed yearly. Paid monthly, the other lines come to about $209 before staging, and the staging plan's monthly price is higher than its $39 yearly rate, so expect at least $248, before the CRM. Set it against the time saved. Five hours a week from drafted chasers alone is about 20 hours a month, worth around $500 at $25 an hour, so the stack pays for itself if the time is used on something useful, such as the stalled files and the vendor calls. If a stage's tool doesn't save a measurable amount in two months, drop it. Some estate agency CRMs already include several of these functions, and Reapit's and Street's pages both list AI features, so before adding a paid tool ask your CRM vendor what its AI covers, using the questions in what to ask your estate agent software vendor.

The order to adopt them in

Start where the work is most frequent and the risk is lowest:

1. Listing drafts from a facts sheet. It happens with every instruction, a person checks the output against a document, and the assistant is already paid for.
2. Progression tracker and drafted chasers. Chasing is the biggest time sink, and the tracker costs nothing but a spreadsheet.
3. Offer summaries. They use the same assistant and the same habit of copying figures exactly.
4. Call answering. Add it once you can show you miss calls, using the free tier first.
5. Viewing feedback summaries. They need the three-question text and a sheet, so they follow once the basics are steady.
6. Valuation prep and staging. Both are worthwhile and both need judgement, so they come last.

Plan for one mistake in particular. A model writing a description from photos and a floor plan can add "moments from the station" to a house that is a long walk away. It slips through when the negotiator proofreads for typos and not for facts, and it surfaces when the first buyer says so at the viewing. The control is the facts sheet rule, plus a habit of reading every particular with the sheet open beside it.

Check that it worked after 60 days with four numbers, compared with the two months before: how long the first reply to a new enquiry takes, the average days from sale agreed to exchange, the number of listings corrected after going live, and staff hours spent on chasers. If the first two improve and the third doesn't worsen, the stack is earning its place.

Questions estate agencies ask when choosing AI tools

Can AI give a valuation?

It can prepare one. An assistant can clean a sold-price export, write the spreadsheet formulas and draft the appointment brief, but the sums run in the sheet and the valuation is the agent's professional opinion, formed after seeing the property. Never quote a figure to a vendor because a tool produced it.

Should AI-staged or AI-edited photos be labelled?

Yes. Label staged images clearly on the listing and keep the original photos in the file, so a buyer is never misled about what the room looks like. Check the portal's rules and your market's advertising rules too, because they differ and change.

What should a small agency buy first?

Start with a shared assistant on a business plan and a one-page facts sheet for each instruction, because listing drafts happen every week and carry little risk. Add progression tracking next, since chasing is the biggest time sink, then call answering if you miss calls. Treat staging and automation tools as later extras.

Can an AI receptionist quote prices or say a property is available?

It can book viewings and take messages. It should state price or availability only from a live CRM field, never from memory, and hand any offer, valuation question or complaint to a person straight away. Test it with calls about a property that has just gone under offer before you let it near real callers.

Further reads

• How Conveyancers Use AI to Cut Admin on Each Transaction — The conveyancer's side of the same transactions
• Estate Agent Complaints Handling: AI-Drafted Replies and a Log — Handle complaints with drafted replies and a log
• Estate Agent AML Checks: Organise Due Diligence Files With AI — Organise due-diligence files without cutting corners
• How Estate Agents Use AI to Write Property Descriptions — A deeper look at writing property descriptions with AI
• AI Virtual Staging for Property Listings: Costs and Disclosure — Staging costs and disclosure in detail
• Can Estate Agents Be Found in AI Search? What to Fix Now — Whether and how agencies show up in AI search
• ChatGPT for Commercial Real Estate: Source-Checked Reports and Comps — A commercial broker's method for ChatGPT market reports, comps summaries and owner emails, with a check on every figure before it reaches a client.
• Using AI for Property Investment: Screening a Deal Step by Step — Screen a rental deal with a traceable input sheet, cautious cash-flow maths and a list of facts to check by hand.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Reapit and Street home pages, Virtual Staging AI pricing page, Smith.ai AI receptionist pricing, Otter, Fireflies and Zapier pricing pages, and the pricing pages for ChatGPT Business, Claude Team and Microsoft 365 Copilot Business, all read on 4 November 2026.