Real Estate AI agent · Tenant & buyer inquiry
Tenant & Buyer Inquiry AI Agent
Restate the published application steps from the system of record, disclose the broker and the automated status, and route the enquiry to the named licensed human, who decides availability, price and who qualifies.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Licensed human
- Answer-only
- After launch
- Agent Care
What this agent does
Answers the question, never decides who qualifies
In
- 01
Receive the enquiry from supported ERP, ILS, MLS or CRM channels, and disclose the broker name and automated status.
- 02
Read the listing file and the published criteria sheet, and carry each answer with the field it came from.
Reason
- 03
Restate the published, uniform application steps and the documents the criteria sheet requires.
- 04
Apply the protected-class configuration set for the jurisdiction, including any source-of-income rules in force there.
- 05
Refuse any figure the record does not hold, rather than inferring one from comparable units.
Decide
- 06
Answer with the same criteria disclosure the listing file carries, whoever is asking.
- 07
Route the enquiry to the licensed property manager or broker of record named on the listing file.
Out
- 08
Retain the verbatim transcript, the disclosures shown and the routing outcome against the enquiry.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent answers questions; the named licensed property manager or broker of record decides who qualifies, and the company stays responsible for what was said.
Example workflow
One enquiry, message to routing
- 1Enquiry receivedRental portal, listing site, brokerage web form or CRM message thread
- 2Record readListing file, published criteria sheet, broker disclosure and jurisdiction configuration
- 3Reply draftedDisclosures, restated steps, refusals and confidence
- 4Controls appliedHeld-act checks, jurisdiction configuration, uniform-disclosure checks and confidence threshold
No human action required
Stages 1 to 4 run unaided, and none of them is authorised to assert availability, price or eligibility — the licensed human's lane opens at the confidence gate.
Goes to the named licensed human.
Adds a supervising-broker read first.
Licensed review
The reply is held with the transcript, its refusals and the confidence.
Send · Amend · Escalate to broker
Approved — sent to the enquirer
- 6CRM and property systems updatedOnly where write access and approval policy allow it
- 7Outcome evaluatedPaired-transcript divergence, refusal outcomes, routing time and complaints
Amendments
A wrong answer is re-run by hand against the criteria on file, and the amendment is counted.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Declaring a unit available or unavailable.
- Approving, denying or waitlisting an applicant.
- Quoting or negotiating rent, deposit, fees or concessions.
- Ordering or interpreting a consumer report.
Automation boundary
Agent acts unaided
- ✓Disclose the broker name and that this is an automated system.
- ✓Restate the published, uniform application steps and documents.
- ✓Route the enquiry to the named licensed human.
- ✓Log the verbatim transcript to the audit store.
Any write happens inside the boundaries agreed at implementation, never ahead of licensed review.
- Issuing an adverse-action notice.
- Characterising a neighbourhood, school or safety condition.
- Deciding a reasonable-accommodation request.
- Choosing which inventory an enquirer is shown.
Example output
One enquiry reply, annotated
Each line the agent sends carries the record field it was read from, because the transcript is the evidence.
Enquiry output · single conversation
Illustrative example- Enquiry
- Two-bed rental
- Restated line
- Applications are taken in the order completed, from the criteria on file
- Read from
- Criteria sheet
- Record read
- Listing file · Yardi
- Confidence
- 91%
- Disclosure
- Broker name and automated status
As received
Taken from the listing file and the criteria sheet — nothing on this side is written by the agent.
Record fields used
- Published criteria sheet
- Listing file status
- Broker disclosure text
Why this answer
It restates a published step; it does not say whether the unit is open or who qualifies.
What the score decides
Below the configured threshold the reply picks up a broker read before it reaches the enquirer.
Action
SendAmendEscalate to brokerValue
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 enquiry
From the enquiry channel
Answer from the record
Draw on the listing file, the published criteria and the jurisdiction rules configured for it.
Take the routine enquiries off the queue
Published process questions arrive already drafted, so the licensed read is the only wait.
Send review where the exposure sits
Refusals and low-confidence replies are marked, so the licensed human's read starts where risk concentrates.
Build an evidence trail
Keep the transcript, the disclosures shown and the routing outcome against the enquiry record.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Property-management ERP
- Yardi Voyager
- RealPage
Rental listings and ILS
- Zillow Rental Manager
- Apartments.com
MLS and IDX data
- CoreLogic Matrix
- MLS Grid
Agent
Tenant & buyer inquiry
- Reads the record
- Answers the question
- Routes to a human
CRM and screening
- Follow Up Boss
- HubSpot
- TransUnion SmartMove
- SafeRent
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 enquirer
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 reply may stand
- L3
- Puts the release in a licensed human's hands
- L4 – L5
- Keep the record and the version honest
- L6
- Pulls answering back when signals degrade
How Nestack evaluates it
Evaluate the whole exchange — not only the last message.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the answer the enquirer reads
- E1Final-output evaluationDid the reply stay inside the held-act boundary?
- E2Step-level evaluationDid the agent read the right listing file, criteria sheet and jurisdiction configuration?
- E3Tool evaluationDid it read and write the correct record and the correct field?
- E4Confidence calibrationDo low-confidence replies actually attract more human amendments?
- E5Slice evaluationHow does performance change across specific enquirer cohorts?
- E6Business outcomeHow many enquiries needed an amendment or a correction after sending?
Floor — the outcome the company 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
- TB-01
Stale availability cache
The listing cache still shows a unit the ERP has already leased.
Stage gathers
Listing file, criteria sheet and jurisdiction config
02 · Reasoning
2 modes
- TB-02
Interpolated rent figure
A null price field is filled from comparable units instead of refused.
- TB-03
Voucher read as disqualifier
Source-of-income status protected in that jurisdiction is read as a screening signal.
Stage proposes
Disclosures, restated steps, refusals and confidence
03 · Tool / write
2 modes
- TB-04
Premature disposition write
A not-qualified status is written before any licensed review.
- TB-05
Unauthorised screening pull
A consumer report is ordered without signed authorisation on file.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- TB-06
Adverse-action notice incomplete
A denial message omits the CRA name and the free-file and dispute rights.
Stage returns
The message the enquirer reads and the record keeps
05 · Change / Version
1 mode
- TB-07
Silent scope regression
An update re-enables neighbourhood summarisation that was switched off.
Stage tracks
Model, prompt, disclosure text and jurisdiction config
- Sev-1 · a held act reaches the enquirer
- Sev-2 · a wrong statement reaches the enquirer
- Sev-3 · source degrades, reply routes to review
Affected slices
One cohort can carry the whole divergence
Voucher-assisted enquirers in source-of-income jurisdictions diverge from their paired counterparts more often than any other cohort, and an aggregate rate will not show it. Nestack reports divergence by slice.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Voucher-assisted enquirers | 4.2% | 3.5× | Review | |
| Accommodation and relay users | 3.0% | 2.5× | Review | |
| Limited-English-proficiency enquirers | 1.8% | 1.5× | Watch | |
| Enquirers with records | 0.6% | 0.5× | Normal |
Bar: paired-transcript divergence-rate lift vs. the all-enquiry baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
A cycle is not closed until a test exists
Nothing closes until the failure exists as a regression case the next release must pass. A sent message cannot be unsent, so detection runs against that suite.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Paired-transcript divergence rises in a cohort.
- 02
Diagnose
The transcript pair, read beside the record it drew on, narrows the cause to one.
- 03
Improve
Whatever changes — config, prompt or disclosure text — is version-stamped and linked to the pair that exposed it.
- 04
Verify
The affected regression cases run again, and a fail stops the release.
- 05
Learn
It becomes a standing test, and the criteria sheet changes with it.
Learn → Detect
The return edge. The next detection 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, enquiry workflow, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Enquiry workflow discovery and boundary definition.Week 1
- 02ERP, ILS, MLS and CRM source assessment.Week 1, Week 2
- 03Disclosure text and jurisdiction protected-class mapping.Week 1, Week 2
- 04Listing-file ingestion and field normalisation.Week 2
- 05Answering logic and record-field binding.Week 2, Week 3
- 06Confidence scoring and refusal routing.Week 3
- 07Licensed review workflow.Week 3, Week 4
- 08CRM and property-system integration.Week 5
- 09Paired-enquiry regression cases.Week 4, Week 5
- 10Guardrails and approval controls.Week 4
- 11Transcript retention 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 property, one channel | ProductionProduction enquiry channels | AdvancedMulti-brand portals |
|---|---|---|---|
| Introduced at Pilot | |||
| Answering from your record and rules | ✓ | ✓ | ✓ |
| Licensed review | ✓ | ✓ | ✓ |
| Answer-quality baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Divergence reporting by cohort | — | ✓ | ✓ |
| Review workflow in your systems | — | ✓ | ✓ |
| Approved CRM write-back | — | ✓ | ✓ |
| Verbatim transcript retention | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-state protected-class rules | — | — | ✓ |
| Approval chains across offices | — | — | ✓ |
| High enquiry volume | — | — | ✓ |
| Enterprise access 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
01The listing files and criteria sheets you publish from
→Listing-file ingestion and criteria mapping
Week 102Representative enquiry transcripts
→Answering baseline, disclosure extraction and record-field binding
Week 203Your broker disclosure and required notices
→Disclosure text and jurisdiction protected-class mapping
Week 104Access to relevant APIs, feeds or exports
→ERP, ILS, MLS and CRM assessment, then integration setup
Week 205Enquiries you would not want repeated
→Paired-enquiry cases and failure-mode testing
Week 406Where a reply must stop and wait for a person
→Confidence scoring, refusal routing, guardrails and approval controls
Week 307Named licensed humans to review replies
→Licensed review workflow, then pilot queue 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
W1Where enquiries arrive, what may be answered, and the boundary around it
W2Listing and criteria sources connected, then a baseline
W3The answering path, its confidence gate and who reviews
W4Paired-enquiry testing, disclosure coverage and the evaluation suite
W5One channel connected, a pilot enquiry queue and targeted corrections
W6A live enquiry queue answered under review, then Agent Care handover
- Reading the band
- Bars are drawn from the workstream list, not rounded to fit. Week 5 genuinely carries two phases.
- At the end of W6
- Verification signed off, and Agent Care picks up monitoring from there.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Real Estate AI agent
Build an inquiry agent around your licensed review chain.
If your enquiry volume already outruns the licensed humans who must answer it, we'll map the exchange, set the automation boundary and name what stays with the broker of record.
- Typical delivery
- 4–6 weeks
- Deployment
- Your ERP and CRM
- Licensed human
- Answer-only
- After launch
- Agent Care available
Nestack Agents · Tenant & buyer inquiry
AGT-RE-02 · Agent Care available after launch