> For the complete documentation index, see [llms.txt](https://thecontentforge.gitbook.io/thecontentforge-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://thecontentforge.gitbook.io/thecontentforge-docs/feature-glossary/glossary/real-estate-operations.md). # Real Estate: Team Operations, Automations and Client-Facing Pages *Last updated: 2026-08-07.* Everything that runs a real estate business around the deals themselves: how a team divides work and signs things off, how the automated assistants operate under supervision, how the workspace is set up, and the pages clients and vendors open from a link without an account. A solo agent needs none of the team setup, and nothing stops working if it is left empty. ## Team Operations One screen for the questions a team lead asks in the same ten minutes: what is waiting, what is late, who has room, and why each lead went where it did. * **Overview** - Brokerage counts for the last thirty days beside a per-person table for the selected team: offices, teams and agents, approvals waiting, service targets missed or at risk, unassigned leads, overdue queue items, median first response time per person, and licences expiring within sixty days. The brokerage half narrows to your own office when that is as far as your access reaches, every figure is a count of records somebody can open, and there is no composite score and no ranking of people against each other. * **Queues** - Shared work lists for unassigned leads, compliance review, marketing approval, transaction review, document review, client follow-up, maintenance, showing requests, referrals and past clients, plus custom queues of your own. Anyone with access can claim an item, which holds it for them until they mark it complete, and anything past its due time is marked late. * **Approvals** - Your inbox of work waiting on your sign-off, showing who submitted it, which step of the review it has reached, and what earlier reviewers said and why. You can approve, ask for changes, or reject outright, anything other than an approval needs a written reason, and you cannot approve something you submitted. An approval covers the exact version reviewed, so a material edit sends it back for sign-off. * **Compliance** - Everything the automatic fair-housing and claims check has flagged as either "do not publish" or "needs a human", with the wording that tripped it, safer phrasing to use instead, and the fix required. You can expand any item to read exactly what was checked. * **Service levels** - Response-time targets for first reply to a lead, showing requests, client messages, document review, offer responses, maintenance requests, compliance review and approval decisions, each with a warning point, an escalation point, an option to count working hours only, and a choice of one team or everybody. A clock only stops when something actually reached the client, so an internal note cannot satisfy a target, and live counts of running, at-risk and missed clocks sit above the list. * **Workload** - Who has room, who is near their limit, who is at it and who is unavailable, with live leads, live deals, open tasks, overdue tasks and running clocks for each person against any caps they have. Rebalancing can be suggested and it proposes only; nothing moves until a person moves it, and the reason is recorded with the lead. * **Lead routing** - Rules that decide which agent a new lead reaches, by taking turns, by agreed share, by who is carrying the least, by first to claim, by specialty match, by service territory, or by sending it to a queue for a person to assign. Each rule also sets what happens when nobody qualifies, whether the lead is assigned outright or offered with a time limit to accept, which markets, property types and languages it applies to, and the order rules run in. * **Routing decisions** - Every routing run is kept with its working: who it went to, the reasons that chose them, whether a fallback was used, and an expandable list of everyone not chosen with the reason for each. A rule that names a protected characteristic such as race, religion, familial status or disability cannot be created at all, so it is never a warning somebody can click past. * **Coaching** - Observations about a person's own work with the supporting evidence and the comparison used, addressed to the person doing it. Only that person can respond, and it is stated plainly that these are not performance ratings and not disciplinary evidence. ## Automations Assistants that watch for work that needs doing, propose a plan, and hold anything consequential until a person approves it. Nothing runs until you switch it on, everything starts in a rehearsal mode where the assistant plans fully and does nothing, and any step that would reach somebody outside the workspace waits for a human decision. * **Today** - The morning picture: how many approvals are waiting on you, what is running, how many problems are open and what today has cost. Critical items and things worth a look are listed with the reason each was raised, and work finished since yesterday opens straight to its full record. * **Recommendations** - Everything the assistants have proposed or finished recently, each with a one-line summary. Marking an item useful or not useful is recorded, and "not useful" stops the same suggestion coming back unless the facts change. * **Approvals** - Work an assistant has prepared and is holding for you. Each entry shows how exposed it is, whether any step cannot be undone, the estimated cost, and when the request expires. * **Approval review** - Before you can approve anything you are shown what triggered the run, what the assistant observed, each step it plans to take with its risk level, which records would change, exactly who outside the workspace would be contacted, why approval is needed at all, and what happens if you approve versus reject. You can approve everything, approve only the steps you tick, or reject with a reason, which is required, and if the plan changes afterwards it comes back to you. * **Active runs** - Everything currently queued, planning, waiting on policy or approval, executing or verifying, with a step counter and the cost so far. Any of them can be stopped mid-flight. * **History** - The full record of past runs with their outcome, cost and any error, each opening a timeline of the plan and its versions, the policy decisions taken, every action with its timing, and the separate check on whether it actually worked. A run that reported success but could not be confirmed is shown as its own distinct outcome rather than as done, and a finished run can be undone where its steps are reversible. Each timeline also takes a rating and a comment, which changes what the assistant suggests next and never changes a safety rule. * **Incidents** - Problems raised automatically and named in plain language: a policy broken, a document that tried to give instructions, a budget exceeded, something that reported success but did not happen, data that nearly went somewhere it should not. Each opens against the run that caused it and can be closed as resolved or as a false alarm. * **Agents** - Turn each assistant on or off and choose how far it may go, from observing only, through recommending and drafting, to doing internal work and acting with your approval. Each one lists what it will and will not do and which actions it can take, rehearsal mode is on until you deliberately switch it off, and the ceiling on each assistant cannot be raised from this screen. * **Available assistants** - Twelve of them: daily command brief, lead response, transaction risk, document chase, marketing operations, seller updates, buyer advisory, property management operations, post-close relationship, compliance and approval, brokerage operations, and local data source health. Each has a stated purpose, a fixed ceiling on how far it may act, and a fixed ceiling on how risky an action it may reach, and none of the ceilings goes beyond acting with your approval. * **Policies** - The safety rules applied to every step before it runs, listed per assistant alongside the highest autonomy and risk level each may reach. These are fixed: a workspace can make an assistant more restricted, never less. * **Tools** - The complete list of actions any assistant is capable of taking, grouped from read-only through to sensitive human-only, each showing how its success gets confirmed and whether it can be undone. Anything not on this list does not exist as far as an assistant is concerned, and an action listed but not granted to a particular assistant is equally out of reach. * **Budgets** - Spend and run counts against any cost or run ceiling in force for the period, per assistant or across all of them, with a clear marker when one is exhausted. Where no budget has been set, an assistant is still bounded by the limits on how much any single run may do. * **Evaluations** - Score an assistant against its own record rather than its own account of itself, and see whether it would be reasonable to give it more freedom. Any unsafe action at all holds it at recommend-only, whatever else it scored. * **Stop everything** - One control that switches every assistant off and cancels every live run, with a confirmation step. Nothing restarts on its own afterwards. ## Settings #### Professional voice and client communication * **Professional Voice and Client Communication** - A twelve-step setup that teaches the system to write as you: your role, your markets and property types, the client types you work with and what each needs to hear, what you want to be known for, your tone across a set of sliders and style questions, how you behave in different situations such as a listing post versus a pricing conversation, your local phrasing and the clichés to avoid, the claims you will not make, how often you want a call to action, and your own writing. The result is one identity that speaks differently to a first-time buyer than to a luxury seller while staying recognisably you. You can jump between steps in any order, and progress is saved as you go. * **Writing samples** - Paste things you have already written, tell the system what each one is, and label how well it represents you so the strong examples carry more weight than the weak ones. You are asked to strip client details first, because this is about your voice and not their information. * **Voice previews and sign-off** - Preview the exact instruction the writer receives, then generate eight sample pieces across different situations and approve or reject each one separately, retrying any that miss. Compliance problems are flagged on the samples themselves, and setup only completes when all eight are approved. * **Always-on claim rules** - Fair-housing and claim safety sit above your preferences with no switch to turn them off, because a preference cannot authorise an unlawful claim. Your own stricter boundaries can be added on top and only ever tighten things further. #### Brokerage * **Offices** - Add each office with its location, timezone, the brokerage name exactly as it must appear, and the licence number, and see how many teams and people sit under each. Office templates inherit the brokerage name and licence number, so they are entered once rather than retyped on every piece. * **Teams** - Create a sales, buyer, listing, luxury, investor, commercial, leasing, property management, marketing, transaction coordination or custom team under an office, optionally naming its lead. Creating one adds the lead and three starting work queues so the team is usable immediately. * **People** - A searchable directory of everyone in the brokerage with their roles, teams, current capacity, lead and deal caps, and licence details including expiry. Licence status records only what was entered or confirmed inside the workspace, and this is stated on the page, because nothing here checks a licensing board. * **Roles** - Sixteen named roles from broker or owner through managing broker, team lead, agent, transaction coordinator and compliance reviewer down to assistant, each with a plain description of what it covers, plus a custom role that grants nothing until capabilities are added one at a time. A role can be granted for one office or across the whole brokerage. * **Offboarding** - Before anyone leaves you see exactly what they hold, including open tasks and approvals waiting on them, then choose who inherits their work or send it to a queue. Nothing is deleted: their past assignments, authorship and approvals stay on the record, and you are told how many assignments moved and how many scheduled posts were paused. * **Permissions** - Grant or deny any of twenty-two individual capabilities, from seeing contacts and commission figures to publishing, approving and administering, each at a chosen scope from nothing up to the whole brokerage, with the reason recorded. A denial overrides whatever the role would allow, you cannot grant yourself a role or one that reaches further than your own, and the page shows the resolved answer for what a person can actually do rather than what they were nominally given. * **Voice and branding** - Layered writing rules that flow from brokerage to office to team to individual: text that must always appear, wording that must never appear, and sections a narrower level is not allowed to change. A narrower level can add its own rules and refine wording left open, but can never remove a lock, and fair-housing rules sit above all of it. #### Importing existing data * **Bulk import from spreadsheets** - Bring in properties, contacts or transactions from a spreadsheet, up to a thousand rows at a time. Column headings are matched by their common names so a file exported from another system usually works untouched, and the first five rows are previewed exactly as they will be read. Each row imports independently so one bad row does not hide the rest, and you get a count of what succeeded and failed with the first twenty failures listed by row number. ## Client-facing pages Pages that people outside the workspace open from a link, with no account and no sign-up. Each one is deliberately narrow: it shows the person only what they need and never reveals anything about other clients, other listings, or the workspace behind it. * **Public property page** - A shareable listing page with photos, description, features, confirmed details, walkthrough links and any required media disclosures, plus an enquiry form that creates a lead in the workspace. A rental listing asks about move-in date, occupants and pets and offers rental-shaped calls to action, enquiries can be switched off once a listing is no longer taking interest, and automated junk submissions are kept out of your leads. * **Open house sign-in** - A phone-friendly sign-in sheet for an open house or rental viewing that asks for contact details, whether the visitor is already working with an agent, their thoughts on the property, and only the housing-related questions the agent has set. It shows nothing about other visitors, the marketing opt-in is separate from signing in and starts unticked with the exact wording stored as the record of what was agreed, and a link that has expired or been closed simply reports itself unavailable. * **Client portal** - A read-only page a seller, buyer, investor or landlord opens to see where things stand: the current stage, confirmed dates, progress checklist, offers received, showing and open house activity, photos with their disclosures, property details, and lists of documents and reports available on request. Figures that have not been confirmed against an official source are labelled as such rather than quietly shown alongside verified ones, the page says plainly that only contract-confirmed dates appear, and the link can be withdrawn at any time or set to stop working on a date the client is shown. * **Showing and service feedback** - A private single-use form for asking a client how a viewing, closing, letting, renewal, management period or annual review went, with an overall rating that is the only required answer, two more on the service and on being kept informed, and an open comment. It never asks for a public review, low scores are as easy to give as high ones with no pre-selected rating, and declining is a visible button rather than an abandoned page. Permission to be quoted later is asked separately, starts switched off, and is explicitly not permission to publish. * **Testimonial permission** - Shows the client the exact words being asked about, then asks how they want it attributed, either anonymously or under a name they type themselves, which places it may be used across social, website, print and video, and whether it may be shortened. Granting requires ticking at least one place, so a yes is never a blanket yes. Nothing is published until they say yes, the permission runs for two years and can be withdrawn, and saying no is a plain button rather than fine print. * **Media upload link** - A link an agent sends to a seller, photographer, contractor or tenant so they can hand over photos and video without an account, which matters because the people holding the actual listing photos are the least likely to sign up for the agent's software. The page asks plain-language rights questions, including who took the files, whether there is permission to share them, whether they may be used for marketing or edited, whether a credit line is required, whether they came from another agent's listing, and whether anything is digitally staged, with "not sure" allowed everywhere except the answers that establish who actually holds the rights. Every link is bounded by an expiry, a file count and a per-file size limit, can be revoked at any time, and the page states that nothing is published automatically. ## Document and media handling What happens to a file after it arrives, whether an agent uploaded it or somebody sent it in through a link. The consistent rule is that nothing becomes a published fact or a published photo without a person deciding it should. * **Document intake** - Every document is stored privately, and the same file arriving twice is reported as a duplicate rather than filed again. Whatever it turns out to be, it lands marked as needing review and not usable in public material. * **Automatic reading and sorting** - Documents are sorted into a type, their text is read page by page including scanned pages so a later citation still points at a real page, and their embedded pictures are lifted out. When the automatic sort is unsure a person can correct the type, and correcting it flags anything already accepted for a second look rather than silently rewriting history. * **Facts with citations** - Figures and named parties are pulled out along with the quote and page they came from, and where two documents disagree the disagreement is flagged rather than quietly resolved. Each extracted fact carries a confidence read you can check for yourself against the cited page. * **Review before anything becomes a property fact** - Extracted information sits as evidence until somebody accepts, edits, rejects it or marks it as needing more information, and resolving a disagreement between two documents is a separate decision from accepting a value. A rejection is final, an edit keeps what the page actually said alongside the correction, and accepting a value never silently overwrites an existing one. * **Suggested follow-ups** - Deadlines and tasks implied by a document are offered as suggestions you can accept or reject, and regenerating the suggestions never disturbs one already accepted or duplicates the task it created. * **Redaction planning** - Before a document is shared or exported you get a reviewable plan of what would be hidden plus a preview, kept so an export can be explained months later. The plan never stores the values it hides, and a document that would be meaningless once redacted is refused rather than sent out blank. * **Media intake** - Photos and video are registered against a property with their dimensions, capture data and an exact-duplicate check. A full listing shoot of forty photos can be sent up in one go, and nothing arriving this way can be publishable until rights are answered for separately. * **Media rights** - The one step that can make media publishable. Somebody answers plain questions about who took it and what it may be used for, and the system decides what those answers are worth rather than accepting a claim of "publishable" directly, and the record is kept even when it grants nothing. * **Listing photo reference capture** - Point the workspace at a live listing page and it records which photos exist so gaps can be spotted and rooms tagged, storing a link back to the original rather than a copy of anyone's photograph. Captured references stay internal and are kept off every public surface unless a person takes responsibility for the rights, and listings on sites whose terms do not permit this return nothing at all, which is the expected answer for most portals. * **Images inside documents** - Pictures pulled out of a PDF or archive inherit the rights of the document they came from, not of whoever uploaded it, so photos from another agent's listing sheet or from inspection work stay internal by default. * **On-demand compliance check** - The same fair-housing and claims check that runs automatically on generated content can be called on any copy or media on request. It is described under Compliance on the Properties and the Property Workspace page. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://thecontentforge.gitbook.io/thecontentforge-docs/feature-glossary/glossary/real-estate-operations.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.