Real Estate AI agent · Listing copy
Listing-Copy Generation AI Agent
Draft listing descriptions from the facts already in the listing record, flag phrasing a licensee should weigh, and hold drafts for the named listing licensee, who decides whether anything is published.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Licensee approval
- Pre-publication
- After launch
- Agent Care
What this agent does
Drafts the copy, not the decision to publish
In
- 01
Ingest the listing record, seller disclosures and configured brand rules from supported MLS, CRM or intake sources.
- 02
Normalise field labels and formats, and carry each fact forward with the source it came from.
Reason
- 03
Draft public remarks and the short-form variants the brokerage's channels require.
- 04
Apply the brand, jurisdiction and MLS field rules configured for the brokerage.
- 05
Bind each drafted claim to a value in the record, and mark what the record leaves ambiguous.
Decide
- 06
Flag phrasing that describes a likely occupant rather than the property.
- 07
Route every draft to the named listing licensee for approval.
Out
- 08
Retain the source facts, the draft, the licensee's edits and the approval against the listing.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent proposes copy; the listing licensee decides whether it is published, and the brokerage stays the advertiser of record.
Example workflow
One listing, record to approval
- 1Listing record receivedMLS record, CRM entry, seller disclosure or listing intake form
- 2Facts assembledMeasurements, counts, permit references and disclosures, each with its named source
- 3Copy draftedPublic remarks, channel variants, flagged lines and confidence
- 4Controls appliedSource-fact checks, jurisdiction and MLS field rules, occupant-language flags and confidence threshold
No human action required
Stages 1 to 4 run unaided, and nothing is published at any of them — the agent is drafting, and the licensee's lane opens at the confidence gate.
Goes to the listing licensee to approve.
Adds a broker compliance read first.
Licensee approval
The draft is held with its source facts, its flagged lines and the confidence.
Approve · Edit · Send to broker review
Approved — released to publish
- 6Listing systems updatedOnly where write access and approval policy allow it
- 7Outcome evaluatedEdit distance, flagged-line outcomes, MLS acceptance and post-publication corrections
Edits
Every licensee edit is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Publishing to the MLS, a portal or any advertisement.
- Originating or adjusting a measurement or room count.
- Deciding what must be disclosed about the property.
- Financing, payment, rate or buy-down language.
Automation boundary
Agent acts unaided
- ✓Draft public remarks and channel variants from the listing record.
- ✓Carry each verifiable fact forward with its named source.
- ✓Apply the brokerage's configured style and field rules.
- ✓Flag phrasing the licensee should weigh, and hold the draft for them.
Any write happens inside the boundaries agreed at implementation, never ahead of approval.
- Age-restricted, 55+ or active-adult wording.
- Reusing prior or another broker's listing text.
- Selecting, clearing or licensing photography.
- Changes to style, jurisdiction or approval rules.
Example output
One set of remarks, annotated
Everything the agent drafts is attached to the record it was drawn from.
Listing-copy output · single listing
Illustrative example- Listing
- Three-bed resale
- Drafted line
- Corner lot with a fenced rear yard and a ground-floor bedroom
- Stated area
- 1,840 sq ft
- Source of record
- Appraisal on file
- Confidence
- 92%
- Attribution
- Brokerage name and licence ID
As received
Taken from the listing record and the appraisal on file — nothing on this side is written by the agent.
Source facts used
- Appraisal measurement
- Seller disclosure
- Permit record
Why this wording
It describes the property, not who might live in it — the line the licensee weighs.
What the score decides
Below the configured threshold the draft picks up a broker read before it reaches the licensee.
Action
ApproveEditSend to broker reviewValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every listing
From the listing record
Draft from the record
Draw on the facts on file and the brokerage's configured style and jurisdiction rules.
Cut the drafting time per listing
Routine public remarks and channel variants arrive already written.
Send review to the risky lines
Flagged phrasing and low-confidence drafts are marked, so the licensee's read starts where risk concentrates.
Build an evidence trail
Retain the source facts, the draft, the flags, the licensee's edits and the approval — on both paths.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
MLS and listing data
- RESO Web API
- Matrix
- Flexmls
- Paragon
- Rapattoni
Syndication
- ListHub
- MLS distribution
- Zillow
- Realtor.com
- Homes.com
Brokerage CRM
- Follow Up Boss
- BoldTrail
- Lofty
- Sierra Interactive
Agent
Listing-copy generation
- Reads the record
- Drafts the copy
- Holds for approval
Transactions and forms
- Lone Wolf
- Dotloop
- SkySlope
- Brokermint
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the listing
Each control wraps the one inside it. What a layer does not catch is named in the failure-mode map below it.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a draft may stand
- L3
- Puts the release in a licensee's hands
- L4 – L5
- Keep the record and the version honest
- L6
- Pulls drafting back when signals degrade
How Nestack evaluates it
Evaluate the drafting workflow — not only the finished paragraph.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the copy the market sees
- E1Final-output evaluationDid every verifiable fact in the draft match its source?
- E2Step-level evaluationDid the agent use the right record, style rules and jurisdiction configuration?
- E3Tool evaluationDid it read and write the correct listing and the correct field?
- E4Confidence calibrationDo low-confidence drafts actually attract more licensee edits?
- E5Slice evaluationHow does performance change across specific listing types?
- E6Business outcomeHow many drafts needed a licensee edit or a correction after publication?
Floor — the outcome the brokerage answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, placed at the stage each one originates.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- LC-03
Non-authoritative source
Area or permit status read from a superseded record.
Stage gathers
Listing record, disclosures, permits and brand rules
02 · Reasoning
2 modes
- LC-04
Occupant-directed phrasing
A property attribute is rewritten as a claim about who would live there.
- LC-06
Over-specified claim
An ambiguous source becomes a definite assertion about permits.
Stage proposes
Public remarks, channel variants and confidence
03 · Tool / write
2 modes
- LC-02
Prohibited-field content
Copy written to a field whose MLS rules exclude it.
- LC-05
Prior-remarks carry-forward
Earlier listing text is reused with its rights and phrasing.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- LC-01
Attribution omitted
Required brokerage name or licence identifier is missing.
Stage returns
The copy the licensee approves and the market sees
05 · Change / Version
1 mode
- LC-07
Silent style regression
A model or rule change widens what the agent will assert.
Stage tracks
Model, prompt, style rules and jurisdiction config
- Sev-1 · copy published outside the boundary
- Sev-2 · a wrong statement reaches the listing
- Sev-3 · source degrades, draft routes to review
Affected slices
Overall quality can hide one bad cohort
Aggregate edit rates can look acceptable while a small number of listing cohorts absorb most of the rework. Nestack reports the licensee-edit rate by slice, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Relists from listing history | 6.1% | 3.4× | Review | |
| Multi-unit rentals | 4.0% | 2.2× | Review | |
| New-construction pre-sale | 2.9% | 1.6× | Watch | |
| Standard resale listings | 1.3% | 0.7× | Normal |
Bar: licensee-edit-rate lift vs. standard-resale baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
Every cycle ends by adding a test
A cycle closes when the failure has become a regression case the next release has to pass. That suite is what the following detection is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Licensee-edit rate rises in a listing slice.
- 02
Diagnose
A broker reads the drafts and the records behind them until the cause narrows to one.
- 03
Improve
The fix is written against a version, with the edits that motivated it attached.
- 04
Verify
Release is blocked until the affected regression cases pass again.
- 05
Learn
The case joins the permanent suite and the brokerage's style rules.
Learn → Detect
The return edge. Detection next time runs against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, drafting workflow, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Listing workflow discovery and boundary definition.Week 1
- 02MLS, CRM and syndication source assessment.Week 1, Week 2
- 03Brand style, jurisdiction and MLS field-rule mapping.Week 1, Week 2
- 04Listing-record ingestion and fact normalisation.Week 2
- 05Drafting logic and source-fact binding.Week 2, Week 3
- 06Confidence scoring and flagged-line routing.Week 3
- 07Licensee approval workflow.Week 3, Week 4
- 08Listing-system and syndication integration.Week 5
- 09Evaluation suite and regression cases.Week 4, Week 5
- 10Guardrails and approval controls.Week 4
- 11Observability and trace instrumentation.Week 5, Week 6
- 12Deployment, documentation and Agent Care handover.Week 6
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel
Final scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
| Capability✓ in scope · — not at this tier | PilotOne MLS, one office | ProductionProduction listing systems | AdvancedMultiple markets / brands |
|---|---|---|---|
| Introduced at Pilot | |||
| Drafting to your record and rules | ✓ | ✓ | ✓ |
| Licensee approval | ✓ | ✓ | ✓ |
| Baseline evaluation | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Slice reporting by cohort | — | ✓ | ✓ |
| Approval workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Observability and tracing | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-MLS and multi-state rules | — | — | ✓ |
| Multi-stage brokerage approvals | — | — | ✓ |
| High listing volume | — | — | ✓ |
| Enterprise controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
- Separate from build
- Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bring
We build with it
01Your listing intake fields and record structure
→Listing-record ingestion and fact mapping
Week 102Representative published listings
→Drafting baseline, style extraction and source-fact binding
Week 203Your brand style guide and approved phrasing
→Brand style, jurisdiction and MLS field-rule mapping
Week 104Access to relevant APIs, feeds or exports
→MLS, CRM and syndication assessment, then integration setup
Week 205Examples of listings that caused rework
→Evaluation suite, regression cases and failure-mode testing
Week 406Your approval thresholds and write-back boundaries
→Confidence scoring, flag routing, guardrails and approval controls
Week 307Named listing licensees to review drafts
→Licensee approval workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Phases sit on the weeks they actually occupy. Week 5 is doubled because evaluation and pilot run together.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Listing workflow discovery, style mapping and the automation boundary
W2Source integration and the drafting baseline
W3Drafting workflow, confidence logic and approval controls
W4Evaluation suite, prohibited-language checks and failure-mode testing
W5Syndication integration, pilot listings and targeted corrections
W6A live listing drafted and approved, then Agent Care handover
- Reading the band
- A bar covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.
- At the end of W6
- Production validation and verification complete, then Agent Care takes over monitoring.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Real Estate AI agent
Build a listing-copy agent around your brokerage's approval chain.
Show us your listing record, your style guide and who signs off. We'll map the drafting workflow, set the automation boundary and name what stays with the licensee.
- Typical delivery
- 4–6 weeks
- Deployment
- Your MLS and CRM
- Licensee approval
- Pre-publication
- After launch
- Agent Care available
Nestack Agents · Listing-copy generation
AGT-RE-01 · Agent Care available after launch