Real Estate AI agent · Trust accounts

Broker Trust-Account Reconciliation AI Agent

Bring the bank balance, the book balance and the trust liability onto one cut-off date, itemise the differences with their evidence, and hold the pack for the broker, who reviews, signs and dates it.

Typical delivery
4–6 weeks
Deployment
Your stack
Broker signs
Read-only
After launch
Agent Care

What this agent does

Assembles the pack, not the signature on it

In

  1. 01

    A bank feed, a journal and a beneficiary ledger set are pulled for one account from supported sources.

  2. 02

    A deposit lands with a truncated originator string, so it is held as unidentified rather than posted.

Reason

  1. 03

    The journal cash balance, the ledger balances and the reconciled bank balance meet on one date, as Colorado requires.

  2. 04

    A beneficiary ledger in deficit is raised as a shortage, not netted — California deducts no such account.

  3. 05

    A difference carries its evidence and a draft of the corrective action Florida records with the explanation.

Decide

  1. 06

    An unidentified receipt, a deposit in transit or an outstanding cheque is aged from the date it first appeared.

  2. 07

    A reconciliation falls due on the cadence the state sets, and the pack routes to the named broker of record.

Out

  1. 08

    The pack, the source document behind each figure and the sign-off are retained on the state's clock.

  2. 09

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

Product statement

Not one resident's balance against the property ledger: the account against what the brokerage owes everyone it holds money for — and the named broker of record reviews, signs and dates it.

Example workflow

One account, feed to signature

  1. 1Cut-off date reachedBank statement, journal, beneficiary ledgers and the prior signed pack
  2. 2Balances assembledJournal cash balance, ledger cash balances and the reconciled bank balance, each with its source
  3. 3Differences itemisedReconciling items, unidentified receipts, ageing and confidence
  4. 4Controls appliedCut-off checks, negative-ledger checks, signatory and ageing checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is posted or adjusted at any of them — the agent is assembling, and the broker's lane opens at the confidence gate.

5DecisionDivides at the confidence threshold
High confidence

Goes to the broker of record to sign.

Low confidence

Adds a controller read first.

Broker sign-off

The pack is held with its differences, its ageing and the confidence.

Sign · Query · Send to controller review

Signed — dated and filed

  1. 6Reconciliation record updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedUnexplained differences, ageing outcomes, examiner findings and post-signature corrections

Queries

Each broker query is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Adjusting, posting or clearing a trust ledger entry.
    • Moving, instructing or scheduling client money.
    • Deciding that a difference has been explained.
    • Certifying that the account and the liability agree.
  • Automation boundary

    Agent acts unaided

    • ✓Pull the bank, journal and ledger balances for one cut-off date.
    • ✓Itemise each difference and carry the evidence behind it.
    • ✓Age unidentified receipts, deposits in transit and stale items.
    • ✓Flag what the broker should weigh, and hold the pack for them.

    The agent is read-only and never signs; nothing it produces moves client money.

    • Funding a shortage from any account.
    • Filing the Nevada annual accounting or the declaration.
    • Adding, removing or authorising an account signatory.
  • Changes to cut-off, threshold or state-rule configuration.

Example output

One difference, annotated

Everything the agent reports is attached to the document it came from — after signature that pack is what an examiner re-performs.

Reconciliation output · single account

Illustrative example
Account
Sales escrow account
Difference line
Cheque to a beneficiary issued before the cut-off and not yet presented
Age at cut-off
41 days
Source of record
Bank statement image
Confidence
93%
Corrective action
Explanation and correction drafted

As received

Taken from the bank statement and the journal on file — nothing on this side is computed by the agent.

Evidence used

  • Bank statement line
  • Beneficiary ledger
  • Deposit slip on file

Why this is a difference

It cleared after the cut-off, so it sits outstanding — a true shortage is cured from the broker's own funds, not edited away.

What the score decides

Below the configured threshold the pack picks up a controller read before it reaches the broker.

Action

SignQuerySend to controller review

Value

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

Each trust account

From the bank and the ledger

03Assembly

Assemble from the record

Draw on the statements, journal and ledgers on file and the state rules configured for the firm.

01Approved path

Three balances, one date

Routine reconciling items arrive already itemised and aged.

02Human review

Send the read to the differences

Ledgers in deficit, unidentified receipts and stale items are marked, so the broker's read starts where risk concentrates.

04

Build an evidence trail

The difference, the item behind it and the broker who signed stay on the reconciliation.

Integrations

Typical integrations

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

  • Trust and escrow banking

    • Bank feeds
    • BAI2
    • OFX
    • Statement images
    • CSV
  • Brokerage accounting

    • QuickBooks
    • Xero
    • Sage Intacct
    • NetSuite
  • Property management

    • Yardi
    • AppFolio
    • Buildium
    • RealPage

Agent

Trust reconciliation

  • Reads the balances
  • Itemises differences
  • Holds for signature
  • Transactions and trust ledgers

    • Lone Wolf
    • Brokermint
    • SkySlope
    • Dotloop
  • 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 signature

The six read outward to inward. What none of them stops is written into the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeRevert to bank-and-book assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, state-rule and cut-off-configuration changes.Track
L4TraceabilityRecord the source documents, the differences, the ageing and the sign-off time.Record
L3Broker sign-offHold the pack for the named broker; it governs release, not whether the signed figures are right.Gate
L2Policy guardrailsTest the pack against the configured state rules and cut-off; a rule failure returns the pack.Restrict
L1Confidence thresholdsRoute a low-confidence pack to a controller read; the score ranks packs, it does not find shortages.Require review
Model corePack produced — three balances, differences, ageing and confidence
L1 – L2
Test whether a pack may stand
L3
Puts the signature in a broker's hands
L4 – L5
Keep the difference and the item behind it
L6
Reverts to bank-and-book assembly when signals degrade

How Nestack evaluates it

Evaluate the assembly workflow — not only the closing figures.

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

Surface — the pack the broker signs

  1. E1Final-output evaluationDid the three balances carry one cut-off date and the right sources?
  2. E2Step-level evaluationDid the agent use the right account, ledger set and state configuration?
  3. E3Tool evaluationDid it read the correct account and the correct period?
  4. E4Confidence calibrationDo low-confidence packs actually attract more broker queries?
  5. E5Slice evaluationHow does performance change across specific account types?
  6. E6Business outcomeHow many packs needed a query or a correction after signature?

Floor — the outcome the licence answers for

Failure modes

Where each failure originates in the agent

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

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • FF-03

      Feeds pulled on two dates

      Bank and journal are read on different days.

    Stage gathers

    Bank statement, journal, ledgers and prior packs

  2. 02 · Reasoning

    2 modes

    • FF-04

      Mid-month deficit missed

      A ledger goes negative and recovers before cut-off.

    • FF-06

      Difference half-explained

      The movement is described and the correction is not.

    Stage proposes

    Balances, differences, ageing and confidence

  3. 03 · Tool / write

    2 modes

    • FF-02

      Transfer in flight netted

      A move between two accounts nets to nothing in the pack.

    • FF-05

      Stale cheque counted twice

      The re-issue and the original both sit outstanding.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • FF-01

      Signatory set unchecked

      A new cheque signer is not matched to the authorisations.

    Stage returns

    The pack the broker signs and the examiner reads

  5. 05 · Change / Version

    1 mode

    • FF-07

      Silent rule regression

      A model or rule change narrows what is raised as a difference.

    Stage tracks

    Model, prompt, state rules and cut-off config

  • Sev-1 · acts on the trust ledger
  • Sev-2 · a wrong pack reaches the signature
  • Sev-3 · a feed degrades, pack holds
See the complete Real Estate failure-mode catalogue

Affected slices

One account can carry most of the differences

A firm-level three-way agreement rate can read clean while a few accounts carry most of the unexplained differences. Nestack reports the unexplained-difference rate by account, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Multi-bank brokerages8.3%3.6×Review
Property-management pools6.6%2.9×Review
Security-deposit accounts3.7%1.6×Watch
Single-account sales firms1.5%0.7×Normal

Bar: unexplained-difference-rate lift vs. single-account-sales baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on a new case

The loop shuts when the unexplained difference is a regression case. That suite is what the next month reconciled is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Unexplained differences rise in one account type.

  2. 02

    Diagnose

    The difference nobody could name, carried month after month, is traced back to the feed that dropped it.

  3. 03

    Improve

    Any change goes out numbered, with the accounts that caused it attached.

  4. 04

    Verify

    One reconciliation case still failing is enough to hold the release back.

  5. 05

    Learn

    One case joins the suite, one line joins the account record.

  6. Learn → Detect

    The return edge. The next month reconciles 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, reconciliation workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Trust workflow discovery and boundary definition.Week 1
  2. 02Bank and ledger source assessment.Week 1, Week 2
  3. 03State rule, cut-off and cadence configuration mapping.Week 1, Week 2
  4. 04Balance ingestion and normalisation.Week 2
  5. 05Three-way logic and evidence binding.Week 2, Week 3
  6. 06Confidence scoring and difference routing.Week 3
  7. 07Broker sign-off workflow.Week 3, Week 4
  8. 08Bank and ledger-system integration.Week 5
  9. 09Difference and ageing cases.Week 4, Week 5
  10. 10Guardrails and signature controls.Week 4
  11. 11Account-trail instrumentation.Week 5, Week 6
  12. 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 tierPilotOne account, one stateProductionProduction bank and ledger feedsAdvancedMultiple states / entities
Introduced at Pilot
Assembly to your accounts and rules✓✓✓
Broker sign-off✓✓✓
Three-way agreement baseline✓✓✓
Introduced at Production
Reporting by account—✓✓
Sign-off workflow in your systems—✓✓
Approved pack write-back—✓✓
Bank-and-ledger integration—✓✓
Introduced at Advanced
Multi-state rule sets——✓
Multi-stage brokerage sign-off——✓
High account volume——✓
Multi-state reconciliation controls——✓
Build priceFrom $5,000From $8,000Custom 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.

  1. 01Your trust accounts and their bank details

    Balance ingestion and account mapping

    Week 1
  2. 02Representative signed reconciliations

    Assembly baseline, ledger mapping and evidence binding

    Week 2
  3. 03Your state rules, cut-off dates and cadences

    State rule, cut-off and cadence mapping, and the boundary

    Week 1
  4. 04Access to relevant feeds, statements or exports

    Bank, ledger and accounting assessment, then integration setup

    Week 2
  5. 05Reconciliations you would not want audited

    Difference cases and the evaluation suite

    Week 4
  6. 06What no signed reconciliation may hide

    Confidence scoring, difference routing, guardrails and sign-off controls

    Week 3
  7. 07Named brokers of record to review packs

    Broker sign-off 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

The chart is drawn in weeks of real work, and that is the reason week five holds a pair of bands.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Trust workflow discovery, state-rule mapping and the automation boundary

W2Source integration and the assembly baseline

W3Reconciliation workflow, confidence logic and sign-off controls

W4Evaluation suite, difference cases and failure-mode testing

W5Ledger integration, pilot accounts and targeted corrections

W6One monthly cycle run under the broker of record, then Agent Care handover

Reading the band
Each bar sits only on the weeks its work is named in. The week 5 overlap is real work, not padding.
At the end of W6
Once the month reconciles, Agent Care assumes the running agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Real Estate AI agent

Build a trust-reconciliation agent around your state's rule.

Show us your trust accounts, your ledger of record and the states you hold money in. The signature, the explanation of each difference and the licence stay with the broker of record.

Typical delivery
4–6 weeks
Deployment
Your accounts and ledgers
Broker signs
Read-only
After launch
Agent Care available

Nestack Agents · Trust reconciliation

AGT-RE-14 · Agent Care available after launch

Build a trust-reconciliation agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your accounts

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.