Five kinds of tool cover most of it: transaction platforms that audit files (SkySlope, Dotloop), a brokerage-level suite such as Rechat, a CRM with AI summaries and follow-up tasks (Follow Up Boss), team assistant seats, and AI coaching or reporting add-ons. Start with the transaction file audit, where a problem found late costs the most.
Published prices start at $20 a seat a month for an assistant, $34.99 a month for an agent's transaction subscription and $340 a month for a transaction platform, while Rechat sells to brokerages by quotation. None of these replaces a transaction coordinator's judgement; they route documents, flag gaps, draft and summarise. I read each vendor's own page for the facts below, and where a vendor publishes no price, the section says so.
The five tool types at a glance, with published costs
Prices are USD list prices from the vendors' own pages in November 2026. "Not published" means the page I read gave no figure.
| Function | What the AI does | Examples | Published cost |
|---|---|---|---|
| Transaction file audit | Routes uploaded pages to checklist items; flags missing signatures or initials | SkySlope; Dotloop | SkySlope from $340 a month, no long-term contract. Dotloop Premium $34.99 per agent a month or $344 a year; Teams and Brokers on request |
| Brokerage-level suite | Marketing design, CRM, deal tracking and an AI assistant in one system | Rechat | Not published; sold through brokerages |
| CRM and lead follow-up | Summaries, drafted messages, suggested tasks, lead prioritisation | Follow Up Boss | Grow $69 per user a month; Pro $499 a month with 10 users; Platform $1,000 with 30 (monthly billing) |
| Team assistant seats | Drafts copy and emails; answers questions from loaded policies | ChatGPT Business; Claude Team; Microsoft 365 Copilot Business | Standard seats $20 a month on annual billing ($25 monthly), from two seats; Copilot Business $21 annual |
| Coaching and reporting | Agent coaching and a spoken digest of brokerage statistics | SkySlope's Ayce and SkySight | Not published |
Transaction file audit: SkySlope and Dotloop
The transaction file is where an AI check earns its keep, because a missing signature or addendum costs the most to fix late. SkySlope's Smart Suite pages describe auto-splitting uploaded PDFs and assigning them to checklist items, and an assist for auditors that detects signatures and initials and flags possible compliance issues. SkySlope's pricing page gives a starting price of $340 a month with no long-term contract, lists DigiSign and Books as add-ons, and does not say which plan includes the AI features. Dotloop's page lists an agent Premium plan at $34.99 a month ($344 a year on annual billing), a free account limited to 10 transactions, and automated compliance workflows for Teams and multi-office compliance for Brokers, with no prices for those tiers and no AI features mentioned.
How to start: pick 20 recently closed files and run them through the audit as if they were live. Compare what it flags with what your coordinator found at the time. Whether it earns its price depends on the sums. Say a brokerage closes 240 transactions a year and a coordinator spends 25 minutes auditing each, 6,000 minutes or 100 hours. If routing and flagging bring that to 15 minutes (an assumption to test), the saving is 40 hours, about $1,200 at $30 an hour, which is less than SkySlope's $4,080 a year starting price. The time saving alone does not pay for it. The case rests on problems caught before closing and on a file that stands up to a broker's audit, so measure those.
The realistic failure is a clean report on an incomplete file. The tool checks what has been uploaded against the checklist you set; a required addendum that was never uploaded does not appear anywhere for it to flag. The remedy is a checklist built for each transaction type that demands the addendum, and a coordinator who signs off the file rather than the report. For the step-by-step version, see compliance file review for brokers, and for which coordinator tasks suit automation, the transaction coordinator task list.
Here is how that looks on one file, illustrative and shortened:
| Checklist item | In the file? | Tool result | Coordinator's action |
|---|---|---|---|
| Purchase agreement | Yes, 9 pages | Routed; initials missing on pages 3 and 5 | Return to the agent for initials |
| Seller disclosure form | Yes | Routed; signed | Tick |
| Inspection addendum | No | Nothing to flag, because nothing was uploaded | Chase the agent; the item stays open |
| Agency disclosure | Yes | Routed; buyer signature missing | Return to the agent |
| Closing instructions | Yes | Routed; signed | Tick |
| Commission statement | Yes | Routed | Tick after checking the figures against the agreement |
The tool caught two missing initials and one missing signature. The missing addendum surfaced only because the checklist listed it and the coordinator saw the item still open, which is the whole argument for keeping a person on the sign-off.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowRechat: the brokerage-level suite
Rechat is the odd one out because it is built for brokerages rather than individual agents. Its site says it partners directly with brokerages for purchasing, and it presents four modules that can be adopted separately: Studio for marketing design, email and social content and listing websites; People, a CRM with contacts, leads, tasks and automation; Deals for deal tracking, listings, checklists, contracts and DocuSign-integrated e-signature; and Lucy, the AI assistant, which the page says can answer questions, draft emails, create campaigns and manage listings. It also mentions back-office tools for admins and says a brokerage can start with one area, such as the CRM alone. No prices are published, so everything below the headline needs a written quote.
Because it replaces several tools at once, it is a firm-wide decision and deserves a firm-wide test. A sensible small-brokerage route: choose the single module that hurts most today, ask for a quote for that module only, and run it with four agents for 60 days. Ask in writing for the price per agent per month, the minimum term, what happens to your contact data if you leave, and exactly which of Lucy's actions need a human to click send. What Lucy does and what it costs goes deeper on the assistant itself. The watch-out is lock-in: the more modules you adopt, the more of your agents' contacts and deals live in one vendor's system, so test the data export before you sign, not after.
CRM and lead follow-up: Follow Up Boss as the reference point
For agents, the CRM is where most daily AI use happens. Follow Up Boss publishes its plans and AI features clearly, which makes it a useful reference even if you choose another system. On monthly billing, Grow is $69 per user, Pro is $499 a month with 10 users included (extra users $49 each), and Platform is $1,000 a month with 30 users (extra $20 each); annual billing takes two months off. Pro and Platform list smart summaries, smart messages, suggested tasks and predictive lead prioritisation. Grow says AI features are included but that some depend on the paid calling add-on, at $39 per user a month.
The sums for 12 agents: Pro for 10 users plus two extra on annual billing is $416 + 2 x $41 = $498 a month, $5,976 a year, about $41.50 per agent a month. How to start: switch on summaries and suggested tasks first, because they save reading time and risk little, and watch lead prioritisation for a month before anyone relies on it. Messages are the watch-out. A drafted text or email still needs the right consent for that channel and a person's eyes, and if you add AI voice calling, the rules on automated calls apply. The comparison of AI CRMs for real estate agents sets Follow Up Boss beside Lofty and BoldTrail.
Team assistant seats for agent support
A team seat is the cheapest tool here and the one agents will use most. ChatGPT Business lists Standard seats at $20 a month on annual billing ($25 monthly) for teams of 2 to 200. Claude Team lists the same $20 and $25 for 2 to 150 seats. Microsoft 365 Copilot Business is $21 per user a month on annual billing (an $18 promotional price runs to 31 December 2026) and is sold as an add-on to a Microsoft 365 business plan. Both OpenAI's and Anthropic's pages state that business content is not used for model training by default. For 12 agents the arithmetic is 12 x $20 x 12 = $2,880 a year on annual billing.
Three jobs justify the seats:
- Policy and procedure answers. Load the brokerage's manual into a Project (ChatGPT Projects and Claude Projects both do this; OpenAI's custom GPTs are scheduled to stop running on 11 December 2026) and let agents ask it questions. Test it first with 20 questions whose answers you know, and require 19 right with the manual's section quoted. A wrong answer on a compliance topic means fixing the manual or removing that topic.
- Drafts agents then own. Listing descriptions, client emails and market updates, each read and edited by the agent. The brokerage-wide rules for listing copy are in the AI use policy below, and a one-week ChatGPT training plan for a team shows how to teach them.
- Offer and contract summaries. A table of an offer's key terms and dates, checked line by line against the document. It is a time saver for the agent and never advice to the client; AI contract review for deadlines and contingencies covers what to verify.
The policy Project needs strict instructions. Something like this works as a starting point:
You answer questions from agents using ONLY the brokerage policy manual uploaded to this project.
For every answer, quote the exact sentence you rely on and give its section number.
If the manual does not answer the question, reply "Not in the manual" and name who to ask (the managing broker).
Never answer from general knowledge about real estate rules. Never give legal advice.With a made-up manual, a good exchange runs: an agent asks, "My buyer client wants to view one of our own listings. What do I have to do?" The answer quotes section 6.3, "Before any showing of a brokerage listing to a brokerage buyer client, the agent must give the dual representation disclosure in writing and notify the managing broker the same day", and adds "Not in the manual: whether the disclosure must be repeated for a second showing. Ask the managing broker." To check it, open the manual and confirm the section and the quote exist. The "Not in the manual" line is the behaviour you want. An answer with no quote counts as a failed test question, however sensible it sounds.
A realistic slip shows why the seats come with a policy. An agent in a hurry pastes a buyer's financing documents into a personal account to summarise them. Personal plans have different training and retention defaults from business plans, so the brokerage cannot say where that data went. The policy names which accounts may be used and which details never go into a prompt, so the answer to "was that allowed?" is already written down.
Coaching and reporting add-ons
SkySlope's Smart Suite page also describes two AI extras: Ayce, which it calls AI-powered agent coaching, and SkySight, a spoken digest of brokerage performance statistics. No prices are published for either. Treat them as optional and test them on your own data. Ask for a sample digest built from your last month and check every figure against your books: if it says closings rose 12% and your records show 27 against 24, that is +12.5%, so ask how it rounds and which dates it counts. For Ayce, ask to see coaching output for a sample agent profile and decide whether your agents would act on it. If neither answer is convincing in a demo, they will not be convincing in month three.
Where the line sits: four things that stay with a person
Agent support tools blur the line between helping an agent and acting for one. Hold these four back from automation, whatever a vendor demo suggests:
- Price opinions. An assistant can lay the comparable sales you supply into a table and point out gaps in them. The pricing recommendation is the agent's and the broker's opinion.
- Negotiating positions. A drafted cover note for a counter-offer is fine. Deciding which terms to counter is not.
- Required forms and disclosures. Use the forms your association or regulator supplies. AI can summarise them, and never rewrites the wording of a required form.
- Who sees which marketing. Do not let any tool pick audiences by characteristics that housing anti-discrimination rules protect, and check the audience settings on any ad it builds.
What a 12-agent brokerage would pay, and what it must save
Take a brokerage with 12 agents closing 240 transactions a year. Using only the published prices above, and leaving out Rechat and the add-ons that have no listed price:
| Item | Basis | Annual cost |
|---|---|---|
| Transaction platform | SkySlope starting price, $340 x 12 | $4,080 |
| CRM | Follow Up Boss Pro, annual billing: $416 + 2 x $41 a month, x 12 | $5,976 |
| Assistant seats | 12 Standard seats x $20 x 12 | $2,880 |
| Total | $12,936 |
That is $1,078 per agent a year, about $90 a month, and $53.90 per transaction at 240 transactions. The question is what it has to buy back. At a value of $40 an hour for an agent's admin time (an assumption), $12,936 equals about 323 hours a year, or roughly 27 hours per agent, which is about 35 minutes a week over 46 working weeks. Whether a CRM that summarises calls, an assistant that drafts emails and a file audit that catches gaps return 35 minutes a week to each agent is something you can measure in a pilot and should not take on trust. The earlier file-audit sum suggested about $1,200 of coordinator time, so most of the case rests on agent hours and on problems caught before closing.
A 90-day adoption plan with measures
Set the measures before the first tool goes live, or the decision at day 90 will rest on impressions.
- Days 1 to 14, baseline. Take the 20 most recent closed files and record the minutes spent auditing each and the number with a missing or unsigned item found after the agent called the file complete (say 6 of 20, which is 30%). List the 15 questions agents ask the broker most often. Record first-response time on the last 30 leads.
- Days 15 to 45, file audit pilot. Run the AI-assisted audit on live files for three agents and the coordinator. Review flags against misses weekly. Sign-off does not change: a person signs every file.
- Days 30 to 60, assistant pilot. Give four agents seats and the signed one-page policy. Run the 20-question test on the policy Project before agents rely on it.
- Days 46 to 75, CRM summaries and tasks. Switch on summaries and suggested tasks for the same four agents. Track first-response time and the share of suggested tasks completed.
- Days 76 to 90, decide. For each tool choose keep, expand or stop against four numbers: late-found missing items (target: none reaching closing), minutes per audit, first-response time, and weekly active agents. Add cost per agent.
Add one stop rule: any tool that lets a compliance-relevant error reach a client is paused until the cause is found and fixed.
The one-page AI use policy to sign before any seat goes live
Six rules fit on a page and cover most of the risk:
- Use approved business accounts only; personal accounts are out.
- No client financial details, identification numbers or bank documents in prompts.
- The agent reads, edits and owns every client message and listing description.
- Listing copy and neighbourhood text are checked for steering and for any description of who lives somewhere; describe the property and the amenities, not the people.
- Edited or generated images are disclosed wherever the rules of the listing platform or your regulator require it.
- The broker may ask to see prompts and drafts at any time.
Policies get read when they are short and tied to a signature. If you would rather work out which of these tools fits your brokerage with someone, that is what my AI implementation consultation covers.
Questions brokers ask before buying AI tools
Which AI tool should a small brokerage buy first?
Start with the transaction file, because a missing signature or addendum costs the most to fix late. Pilot an AI-assisted file audit on 20 files before buying anything for agents, then add CRM summaries, then team assistant seats. Rechat is the one to evaluate as a firm-wide decision, since it is sold to brokerages rather than to individual agents.
Are team assistant seats safe for client details?
Business plans from the main assistant vendors state that they do not train on business content by default, but that covers training, not what an agent should paste. Keep financial details and identification out of prompts, load policies rather than client files, and have every agent sign a one-page AI use policy.
What does an AI stack cost a 12-agent brokerage?
On published prices, a transaction platform from about $340 a month, a CRM near $498 a month and 12 assistant seats at $20 each come to roughly $12,900 a year, about $90 per agent per month, before any brokerage-suite quote or add-ons. Use it as a planning figure and recheck each price before buying.
Further reads
- Best ChatGPT Plan for Realtors: Free, Plus or Business? — Free, Plus or Business: which ChatGPT plan suits an agent or a team.
- Follow Up Boss Pricing: What a Small Team Really Pays — What a small team really pays for Follow Up Boss.
- Follow Up Boss AI: Which Plan Adds Smart Summaries and Suggested Tasks? — Which Follow Up Boss plan adds summaries and suggested tasks.
- Virtual Assistant or AI for a Real Estate Agent: What Goes Where — Which jobs go to a virtual assistant and which to AI.
- AI Real Estate SEO: Neighbourhood Pages That Avoid Steering — Neighbourhood pages that avoid steering when AI writes them.
- AI Disclosure in Real Estate: What Listing Photos and Copy Must Say — What AI-edited photos and AI-written copy must disclose.
- AI for a Real Estate Agency: Roles, Tools and a First-Month Plan — Roles, tools and a first-month plan for a whole agency.
- AI for Rental Properties: Screening Enquiries, Viewings and Tenant Replies — One landlord with six rentals, walked through enquiry replies, viewing scheduling and tenant messages with AI, with numbers and fair-housing guardrails.
- Real Estate Agent Recruiting With AI: Outreach and Onboarding Plan — A four-touch recruiting sequence, an onboarding checklist and a first-30-days plan for a brokerage, with the compliance lines AI must not cross.
- Real Estate Wire Fraud and Voice-Clone Calls: A Call-Back Script — An office call-back procedure for wire instructions, with a staff script, a client notice and a written policy that holds up even against a cloned voice.
- AI Policy for a Real Estate Brokerage: What Agents May and May Not Do — Turn a broker AI policy into rules an agent can follow on a Tuesday: do and don't lists, sample clauses and a sign-off sheet.
- Places to Use AI in Real Estate: Eight Jobs for a Small Brokerage — Prioritise eight practical AI jobs by volume, source quality and risk, then pilot one workflow with a review owner.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: vendor pricing and product pages read in November 2026 (SkySlope, Dotloop, Rechat, Follow Up Boss, OpenAI ChatGPT Business, Anthropic Claude Team, Microsoft 365 Copilot Business). Prices are USD list prices and change, and several vendors publish none. Worked figures are illustrative.