Product
The AI tools, and what each one cannot do
Every tool here drafts something a licensed agent then reviews. This page says what each produces, and is equally specific about where the output stops being trustworthy.
The portal runs a set of AI tools across drafting, research and pipeline work: listing descriptions, outreach and reply drafts, objection scripts, CMA summaries, meeting preparation, transaction risk review and fair housing scanning. Each produces draft material for a licensed agent to review. None of them verifies a fact, and none can substitute for professional judgement on price, disclosure or compliance.
Key takeaways
- Every tool produces a draft. Nothing generated here is reviewed, verified or approved by a licensed professional before it reaches the agent.
- Tools run on more than one model provider, and an administrator can assign a different model per feature rather than being locked to one vendor.
- The fair housing scanner flags language patterns; it is a prompt to look again, not a compliance certification.
- Lead scoring orders a follow-up queue from observed signals. It does not read intent and cannot predict a closing.
What the drafting tools produce
The drafting group turns structured input into first-draft copy. The listing writer takes property parameters and returns marketing copy. The email writer covers a set of common email types with selectable tone. The objection handler returns several distinct response strategies for a stated objection rather than one script.
The useful framing is time-to-first-draft, not quality-versus-a-professional. A generated listing description is a starting point that removes the blank page; it still needs the specifics only the agent who walked the property knows, and it still needs a compliance read before it goes anywhere public.
What the research and pipeline tools estimate
The research group summarises rather than drafts. CMA summarisation condenses comparable data into something a client can read. Meeting preparation assembles what is known about a contact before a call. Transaction risk review reads a deal state and surfaces what commonly goes wrong from there.
These are estimates over data the system already holds. They inherit whatever is wrong or missing in that data, and none of them reaches outside it to verify a figure. A summary that looks confident about a comparable is confident about the record, not about the property.
Lead scoring belongs in this group and deserves the sharpest caveat. It ranks prospects by patterns in data available at a moment, which is useful for ordering a follow-up queue. It cannot verify intent, financing or insurability, and a high score is not a probability of closing.
| Tool group | What it produces | What it cannot do |
|---|---|---|
| Listing and marketing drafts | Property copy, social captions, listing presentations | Verify a property fact or clear fair housing language on its own |
| Outreach and reply drafts | Email and follow-up drafts by type and tone | Know whether a contact consented to be contacted |
| Lead scoring and prioritisation | A relative ranking of a follow-up queue | Read intent, confirm financing, or predict a closing |
| CMA and market summaries | Condensed comparable and market narratives | Establish a defensible valuation or replace an appraisal |
| Fair housing scanning | Flags on language patterns that commonly cause problems | Certify that copy is compliant |
More than one model, assigned per feature
The platform routes to more than one model provider rather than standardising on a single vendor, and an administrator can assign a different model to each feature. Fast, inexpensive models handle high-volume work such as parsing an inbound lead; stronger models handle work where quality matters more than latency.
That mapping is configurable rather than fixed, which matters for two reasons. A feature that produces weak output can be moved to a stronger model without waiting for a release, and a provider outage does not take every AI feature down at once.
Usage is logged per call with the feature and model recorded, and per-user rate limits apply. There is a global switch that disables AI features outright.
Multimodal input, where it earns its place
Lead capture accepts a screenshot as well as text. An agent who receives a messy enquiry by SMS can paste the text or drop in an image, and the parser returns structured lead fields to review before anything is saved.
The review step is the design. Parsed output pre-fills a form rather than writing a record directly, so a misread phone number is corrected by the agent rather than discovered three weeks later in a failed follow-up.
What the AI does not change
Machine-drafted marketing carries the same obligations as anything an agent writes. Fair housing rules, advertising and disclosure requirements, and broker supervision apply identically. A scanner that finds no problem is not a defence.
RealEst Technologies LLC is a software company, not a brokerage, and does not review output before it reaches an agent. Where a draft touches law, tax, insurance or valuation, the licensed professional owns the result.
Common questions
Do I need my own AI API key?
No. Model access is covered by the plan rather than billed per token, so there is no separate provider account to set up and no usage invoice to reconcile.
Which model runs a given tool?
It depends on the feature and is configurable. Fast models handle high-volume parsing work, stronger models handle quality-sensitive drafting, and an administrator can change the assignment per feature.
Is generated listing copy safe to publish as-is?
No. Treat it as a first draft. It has not been checked against the property, against fair housing requirements, or against brokerage advertising rules, and publishing it unreviewed puts those obligations on the agent anyway.
Can lead scoring tell me who is going to buy?
No. It ranks records by resemblance to patterns in past data for whatever outcome it was trained on. It cannot observe intent, and treating a score as a forecast of a closing is the most common way to misuse it.