Property-Tax Appeal Evidence AI Agent

What this agent does

Assembles the record, not the number on it

A notice of value lands, and the parcels it covers are scoped to the roll and the jurisdiction they sit in.

A deadline falls, and it is rebuilt from that year's published source rather than read from a cached table.

Reason

A California parcel is read against the current Board of Equalization letter, not last year's county list.

A Texas notice arrives, and the 30-day clock at Tex. Tax Code §41.44 runs beside the 15 May date.

A parcel's income and expenses are built on the accounting basis the form requires, for the period it requires.

Decide

A parcel's signer is checked against the jurisdiction's eligibility rule and the authorisation on file.

A parcel's comparable set is assembled with its provenance — the opinion of value stays with a named person.

Out

A parcel is decided, and the escalation calendar opens that day, shorter than the filing calendar behind it.

Execute write actions only inside the approval boundaries agreed during implementation.

→ Product statement

The agent assembles the record; a named person signs — in California under penalty of perjury, and states the opinion of value the agent never forms.

Example workflow

One parcel, notice to signature

AgentHuman

1Notice of value receivedAssessment roll, notice of value, prior determination or portfolio intake

2Record assembledIncome and expense lines, comparables and provenance, authorisations, each with its source

3Package proposedExhibit set, open gaps, deadline and confidence

4Controls appliedJurisdiction deadline checks, signer-eligibility tests, cross-filing reconciliation and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is signed or lodged at any of them — the agent is assembling, and the signer lane opens at the confidence gate.

5DecisionTurns at the confidence threshold

High confidence

Goes to the named signer to review.

Low confidence

Adds a tax-counsel read first.

Signer review

The package is held with its exhibits, its open gaps and the confidence.

Approve · Amend · Send to counsel review

Approved — released for signature▼

6Appeal systems updatedOnly where write access and approval policy allow it

7Outcome evaluatedExhibit completeness, on-time filing, dismissals and post-filing amendments

Amendments

Each amendment made in review is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Forming or implying an opinion of value.

Submitting the application to the authority.

Signing or swearing to the application.

Appearing before the notary or the board.

Automation boundaryAgent acts unaided

✓Assemble the exhibit set that the jurisdiction form actually demands.

✓Rebuild each deadline from that year's published source.

✓Reconcile the income lines to the books and to prior filings made.

✓Flag the gap the signer must close, and hold the package.

Writes land inside the boundaries agreed at implementation, never ahead of a human signature.

Making legal argument on the parcel.

Certifying the accounts as an independent CPA.

Agreeing to extend a California hearing date.

Changes to deadline, signer or exhibit rules.

Example output

One parcel package, annotated

Everything the agent assembles stays attached to the record it came from, and stays after the filing.

Appeal-record output · single parcelIllustrative example

Parcel

Assembled line

Filing deadline

Deadline source

Confidence

Signer of record

Alameda County parcel

Comparable set and provenance assembled, opinion-of-value box left blank

15 September 2026

BOE LTA 2026/023

91%

Owner officer, under penalty of perjury

As receivedTaken from the assessment roll and the county's published dates — nothing on this side is written by the agent.

Evidence cited Recorded sales records Operating statement Prior board determination

Why the box is blankSworn under penalty of perjury by a named person, and it is logged.

ActionApproveAmendSend to counsel review

What the score decidesBelow the configured threshold the package picks up a tax-counsel read before.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04

Every parcelFrom the assessment roll

03Assembly

Build from the roll and the form

Ask what the jurisdiction's form actually demands, not what last year's package happened to contain.

01Approved path

The deadline does not move

Routine parcels arrive with their exhibits assembled and their deadline dated.

02Human review

Send review where the year can be lost

Open gaps and contested signers come first — in New Jersey an unanswered Chapter 91 request bars the appeal.

04Build an evidence trail

The parcel, the exhibit behind it and the signer who filed stay on the application.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

Assessment and roll dataCounty assessor rolls
Public parcel and sales data

Property accountingYardi Voyager · MRI
RealPage · Entrata

Appeal and tax systemsitamlink · PTMS
OneSource Property Tax

Agent

Property-tax appeal evidence

Reads the roll
Assembles the record
Holds for signature

Documents and authorisationsPowers of attorney · forms
SharePoint · document stores

Observability & evaluationOpenTelemetry · 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 application

Six rings, outermost first. Whatever clears the innermost is named in the map beneath this.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modePull assembly back to exhibit-only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, deadline-source and form-edition changes.Track

L4TraceabilityRecord the exhibits cited, the gaps, the amendments and the signature time.Record

L3Signer reviewHold packages for the named signer; it governs release, not whether a signed application is right.Gate

L2Policy guardrailsTest packages against jurisdiction deadlines, signer rules and exhibit lists; a failure returns the package.Restrict

L1Confidence thresholdsRoute low-confidence packages to a tax-counsel read before the signer sees them.Require review

Model corePackage assembled — exhibit set, open gaps, deadline and confidence

L1 – L2Test whether a package may stand

L3Puts the signature in a person's hands

L4 – L5Keep the parcel and the exhibit behind it

L6Reverts to exhibit assembly when signals degrade

How Nestack evaluates it

Evaluate the whole appeal season — not only the finished application.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the application the board reads

Depth of coverage ▼

E1Final-output evaluationDid each assembled line trace to a document in the file?

E2Step-level evaluationDid the agent use the right jurisdiction, form edition and deadline source?

E3Tool evaluationDid it read and write the correct parcel and the correct roll year?

E4Confidence calibrationDo low-confidence packages actually attract more signer amendments?

E5Slice evaluationHow does performance change across specific authorities and parcel types?

E6Business outcomeHow many packages needed an amendment, or a correction after the application went in?

Floor — the outcome the owner answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, set at the stage where each one first begins.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

FK-03

Cached deadline table

Deadline read from last year's county list.

Stage gathersRolls, notices, published dates and authorisations

02 · Reasoning2 modes

FK-04

Implied value opinion

A comparable set is arranged so it reads as a recommendation.

FK-06

Wrong accounting basis

Income reported on a basis the return does not use.

Stage proposesExhibit set, open gaps, deadline and confidence

03 · Tool / write2 modes

FK-02

Ineligible signer

Package built for a signer the forum rejects.

FK-05

Cross-filing mismatch

Appeal and RPIE income differ on the same parcel.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

FK-01

Incomplete exhibit set

A required schedule or attachment is missing.

Stage returnsThe application the signer swears and files

05 · Change / Version1 mode

FK-07

Silent deadline drift

A published date changes and the calendar keeps the old one.

Stage tracksModel, prompt, deadline sources and form editions

Sev-1 · a package leaves the boundary Sev-2 · a wrong line reaches the application Sev-3 · date source changes, package to review

See the complete Real Estate failure-mode catalogue →

Affected slices

One taxing authority can carry the exposure

A portfolio-wide on-time filing figure can read well while one taxing authority or parcel type carries most of the incomplete sets. Nestack reports exhibit completeness by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Newly reassessed parcels	9.2%	3.7×	Review
Multi-authority portfolios	6.7%	2.7×	Review
Vacant or partly let parcels	4.3%	1.7×	Watch
Stable let parcels	1.4%	0.6×	Normal

Bar: gap-rate lift vs. stable-parcel baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

Every season ends by adding a case

A cycle is done when the missed deadline has become a case the next release must pass. That suite is what the next season filed is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Exhibit completeness falls in a taxing-authority slice.

02Diagnose

The parcel that was over-assessed for a year nobody appealed is read back until the cause narrows to one.

03Improve

Stamp the change; the parcels behind it are filed against that number.

04Verify

Each touched parcel case is run again, and one red stops the release.

05Learn

It stays a standing test, and the deadline rules move with it.

Learn → DetectThe return edge. Next season 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, appeal workflow, evaluation, integration, then production validation and handover.

Workstream Week 1Week 2Week 3Week 4Week 5Week 6

01Appeal workflow discovery and boundary definition.

02Roll and accounting source assessment.

03Jurisdiction deadline, form and signer mapping across markets.

04Parcel and roll ingestion, normalisation.

05Exhibit assembly and source binding.

06Confidence scoring and gap routing.

07Signer review workflow.

08Assessment and tax-system integration.

09Deadline and exhibit cases.

10Guardrails and signing controls.

11Parcel-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 authority, one portfolio ProductionProduction tax systems AdvancedMultiple states / authorities

Introduced at Pilot

Assembly to your rolls and forms✓✓✓

Signer review hold✓✓✓

Exhibit-completeness baseline✓✓✓

Introduced at Production

Reporting by taxing authority—✓✓

Review workflow in your systems—✓✓

Approved write-back—✓✓

Assessment-data integration—✓✓

Introduced at Advanced

Multi-authority rule sets——✓

Multi-stage signer reviews——✓

High parcel counts——✓

Multi-jurisdiction filing controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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 bringWe build with it

01Your parcel list and roll identifiers → Parcel and roll ingestion and evidence mappingWeek 1

02Representative filed applications → Assembly baseline, source binding and gap detectionWeek 2

03Your jurisdictions and signing conventions → Jurisdiction deadline, form and signer mappingWeek 1

04Access to relevant APIs, feeds or exports → Roll and accounting source assessment, then integration setupWeek 2

05Applications you would not want dismissed → Deadline cases and the evaluation suiteWeek 4

06What no signed application may guess → Confidence scoring, gap routing, guardrails and signing controlsWeek 3

07Named signers to review packages → Signer review workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each phase is as wide as the weeks it consumes, so the doubling at week five is work, not layout.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Appeal workflow discovery, rule mapping and the automation boundary W2Source integration and the assembly baseline W3Package workflow, confidence logic and signing controls W4Evaluation suite, deadline checks and failure-mode testing W5Tax-system integration, pilot parcels and targeted corrections W6One appeal season run under the tax director, then Agent Care handover

Reading the bandA bar sits only on the weeks its work is named in. The week 5 overlap is real work, not padding.

At the end of W6When the season validates, Agent Care takes the agent on.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Real Estate AI agent

Build an appeal-record agent around the person who signs.

Show us your parcels, your jurisdictions and who signs. Bring one taxing authority and one filing season, and we will map the assembly workflow, rebuild that season's deadlines from the published source, and name what stays with a person — starting with the opinion of value, which in California is stated on the face of the application under penalty of perjury and, if the county board does not hear the application within two years, becomes the value taxed under Revenue and Taxation Code §1604(c).

Discuss this agent →

Nestack Agents · Property-tax appeal evidenceAGT-RE-18 · Agent Care available after launch