Main site

Construction AI agent · AP and payroll

AP, Payroll & Back-Office AI Agent

Three instruments decide this workflow — a Statement of Compliance, a lien waiver, a payment schedule — and the agent signs none of them. It computes and reconciles; the named signatory signs.

Typical delivery
4–6 weeks
Deployment
Your stack
Signatory signs
Compute only
After launch
Agent Care

What this agent does

Computes the figure; it signs nothing

In

  1. 01

    Compute the statutory deadline register for payment, certification and lien clocks, and escalate on it to named humans.

  2. 02

    Compute hours, deductions and fringe credits from the approved timesheets.

Reason

  1. 03

    Reconcile hours, classifications and fringe credits against that determination and the source timesheet.

  2. 04

    Reconcile each payment application against the certified value on file and report every difference as an exception.

  3. 05

    Maintain immutable per-subcontractor trust and retention ledgers, with statements and overdraft alerting.

Decide

  1. 06

    Maintain the evidence pack behind every figure — timesheets, rates, the determination and the prior certification.

  2. 07

    Route the pack, the open clocks and every unresolved variance to the signatory named for that instrument.

Out

  1. 08

    Retain every timesheet edit, computation, draft instrument and signature against the run.

  2. 09

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

Product statement

The agent computes and reconciles; the named signatory signs each instrument — and once a deadline passes the way back is jurisdictional: in the UK, true-value adjudication in a later cycle; in Australia, generally none within that cycle.

Example workflow

One payment run, timesheet to signature

  1. 1Payment cycle openedTimesheets, payment applications, wage determinations and the project's configured rules
  2. 2Figures computedHours, classifications, fringe credits, deductions and retention, each against the record it came from
  3. 3Pack assembledDraft instruments, open clocks, unresolved variances and confidence
  4. 4Controls appliedDetermination-currency checks, jurisdiction rule checks, trust-ledger checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is signed, certified or released at any of them — the agent is computing, and the signatory's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the named signatory to sign.

Low confidence

Adds a payroll-manager read first.

Signatory decision

The run is held with its computations, its open clocks and the confidence.

Accept run · Return · Send to payroll review

Signed — recorded against the run

  1. 6Payroll and job-cost systems updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedInstrument timing, variance outcomes, clock acknowledgements and corrections after signature

Returns

Every returned run is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Signing or submitting a Statement of Compliance.
    • Executing a lien waiver, conditional or unconditional.
    • Issuing or withholding a payment schedule or pay-less notice.
    • Classifying a worker or assigning a wage classification.

  • Automation boundary

    Agent acts unaided

    • ✓Compute the statutory deadline register and escalate on it to the named owner.
    • ✓Reconcile hours against the determination.
    • ✓Assemble the evidence pack for the human signatory.
    • ✓Maintain trust and retention ledgers, with statements and overdraft alerting.

    Any write happens inside the boundaries agreed at implementation, never onto an instrument.

    • Certifying, reducing or setting off a payment application.
    • Setting a retention percentage, or moving statutory trust money.
    • Filing, releasing or letting a lien or notice lapse.
  • Determining or overriding a CIS deduction rate.

Example output

One certified line, annotated

Everything the agent computes is attached to the timesheet it was built from.

Payroll output · single worker-week

Illustrative example
Classification
Configured class
What was computed
Base and fringe reconciled against the approved timesheet and the determination configured for the project
Hours reconciled
38.5 hours
Rate source
Configured determination
Confidence
89%
Instrument
Draft — unsigned

As received

Taken from the timesheets and the determination on file — nothing on this side is written by the agent.

What was read

  • Approved timesheet
  • Configured determination
  • Prior certification

Why this figure

Base and fringe reconcile to the timesheet — the signatory still certifies it.

What the score decides

Below the threshold the run picks up a payroll-manager read before the signatory.

Action

Accept runReturnSend to payroll 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

Every instrument

From the timesheet and the determination

03Computation

Compute against the record

Hours, classifications and fringe credits are computed from the approved timesheet and the determination and rules configured for the project.

01Approved path

Meet the clock, not the cycle

Deadline registers, reconciliations and evidence packs are ready before the signatory opens the run.

02Human review

Send review to the hard deadlines

Open statutory clocks and draft instruments are marked, so the signatory's read starts where nothing reopens — a passed deadline, a waiver already signed.

04

Build an evidence trail

The timesheets, rates and determination behind each figure stay with the run.

Integrations

Typical integrations

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

  • Construction ERP and job cost

    • Sage 300 CRE
    • Foundation
    • Trimble Viewpoint

  • Certified payroll

    • LCPtracker
    • Points North Certified Payroll

  • Progress claims and payments

    • Payapps
    • Progressclaim.com

Agent

AP, payroll and back office

  • Reads the timesheets
  • Computes the run
  • Holds for the signatory

  • Time capture and determinations

    • eBacon
    • Procore Workforce
    • Published wage determinations

  • 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 instrument

Each control wraps the one inside it. A layer can return a run; none of them signs, and none reopens a statutory window that has closed.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modePull the agent back to computation only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, determination-source and rounding-default changes.Track
L4TraceabilityRecord the timesheets, the computation, the draft instrument and who signed it.Record
L3Signatory decisionHold runs for the named signatory; it governs release, not whether the signed figure is right.Gate
L2Policy guardrailsTest runs against determination-currency, jurisdiction and trust-ledger rules; a failure returns the run.Restrict
L1Confidence thresholdsRoute low-confidence runs to a payroll-manager read before the signatory sees them.Require review
Model coreRun produced — computations, draft instruments, open clocks and confidence
L1 – L2
Test whether a run may stand
L3
Puts the signature in a named person's hands
L4 – L5
Keep the computation and its inputs reconstructable
L6
Reduces to reconciliation when signals degrade

How Nestack evaluates it

Evaluate the whole payment run — not only the figure on the instrument.

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

Surface — the instrument that leaves the office

  1. E1Final-output evaluationDid every figure reconcile to the timesheet and the determination it cites?
  2. E2Step-level evaluationDid the agent use the current determination and the rules configured for the project?
  3. E3Tool evaluationDid it read and write the correct run, ledger and draft instrument?
  4. E4Confidence calibrationDo low-confidence runs actually attract more returns?
  5. E5Slice evaluationHow does performance change across specific jurisdictions?
  6. E6Business outcomeHow many runs were returned, or corrected after the instrument was signed?

Floor — the deadline that does not reopen

Failure modes

Where each failure originates in the agent

Seven failure modes, grouped under the stage that produces each one.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • PW-01

      Superseded determination read

      A determination no longer in force, or a rule reading since vacated.

    Stage gathers

    Timesheets, applications, determinations and rules

  2. 02 · Reasoning

    2 modes

    • PW-02

      Coverage scoped to the wrong rules

      Work is scoped under a rule set the project no longer runs under.

    • PW-03

      Fringe credit on the wrong base

      An underpayment is produced that reads as compliant on the form.

    Stage proposes

    Computed figures, open clocks and confidence

  3. 03 · Tool / write

    2 modes

    • PW-04

      Unconditional waiver released

      A release goes out on a remittance advice, not on cleared funds.

    • PW-05

      Trust ledger drawn down

      Retention money funds an operating payment run.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • PW-06

      Reasons for withholding omitted

      A schedule states no basis, so there is no valid notice at all.

    Stage returns

    The instrument the signatory signs and issues

  5. 05 · Change / Version

    1 mode

    • PW-07

      Rounding defaults shifted

      A release changes rounding or fringe allocation across every later run.

    Stage tracks

    Model, prompt, determination sources and rounding

  • Sev-1 · a held act, or an unsigned instrument
  • Sev-2 · a wrong figure reaches a signature
  • Sev-3 · a source degrades, the run is held
See the complete Construction failure-mode catalogue

Affected slices

The worst jurisdiction disappears into the total

A month can close on an acceptable instrument rate while one jurisdiction group carries most of the misses, so Nestack reports the missed-or-invalid-instrument rate by slice.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Multi-state US federally funded work4.3%3.6×Review
AU trusts across NSW and QLD3.6%3.0×Review
UK Construction Act work with CIS1.8%1.5×Watch
Single-jurisdiction monthly cycles0.6%0.5×Normal

Bar: missed-or-invalid-instrument lift vs. an all-instrument baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The suite is what makes a cycle finished

The cycle closes on a test the next release has to pass, not on a write-up of what went wrong in the last one.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Missed-or-invalid-instrument rate rises in a jurisdiction slice.

  2. 02

    Diagnose

    Where the misses cluster on one clock or one determination, the cause is there; where they do not, the run is read end to end.

  3. 03

    Improve

    Each change is version-stamped and linked to the run that surfaced it.

  4. 04

    Verify

    The release waits on the affected cases passing again.

  5. 05

    Learn

    It becomes a permanent case and a change to the deadline register.

  6. Learn → Detect

    The return edge. The suite the next detection meets is longer by this one.

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, payment workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Payment workflow discovery and boundary definition.Week 1
  2. 02Timesheet and determination assessment.Week 1, Week 2
  3. 03Jurisdiction, determination and trust-rule mapping.Week 1, Week 2
  4. 04Timesheet ingestion and hours normalisation.Week 2
  5. 05Reconciliation logic and deadline computation.Week 2, Week 3
  6. 06Confidence scoring and variance routing.Week 3
  7. 07Signatory decision workflow.Week 3, Week 4
  8. 08Payroll and job-cost integration.Week 5
  9. 09Instrument-timing regression cases.Week 4, Week 5
  10. 10Guardrails and signature controls.Week 4
  11. 11Computation-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 project, one jurisdictionProductionProduction payroll systemsAdvancedMultiple projects / regimes
Introduced at Pilot
Reconciliation and clock register✓✓✓
Signatory decision✓✓✓
Instrument-timing baseline✓✓✓
Introduced at Production
Reporting by jurisdiction—✓✓
Sign-off workflow in your systems—✓✓
Approved write-back—✓✓
Finance-system integration—✓✓
Introduced at Advanced
Mixed prevailing-wage regimes——✓
Multi-stage sign-off——✓
High payment volume——✓
Enterprise trust controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, payment 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 timesheet sources and pay rules

    Timesheet ingestion and reconciliation mapping

    Week 1

  2. 02Representative signed payment runs

    Reconciliation baseline, fringe logic and variance rules

    Week 2

  3. 03Your determinations, trust and retention rules

    Jurisdiction and trust-rule mapping, and automation-boundary definition

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Timesheet, determination and claims assessment, then integration setup

    Week 2

  5. 05Cycles where an instrument went out wrong

    Instrument-timing cases and failure-mode testing

    Week 4

  6. 06The clocks that must never be met by the agent alone

    Confidence scoring, variance routing, guardrails and approval controls

    Week 3

  7. 07Named signatories for each instrument

    Signatory decision 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

A phase is drawn only over the weeks it is worked. Evaluation and pilot share week 5.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Payment workflow discovery, jurisdiction mapping and the boundary

W2Source integration and the reconciliation baseline

W3Deadline logic, confidence scoring and sign-off controls

W4Evaluation suite, guardrails and failure-mode testing

W5Payroll-system integration, a pilot run and targeted corrections

W6One full payment and certification run under the signatory

Reading the band
Each bar ends where its named work ends. The week 5 overlap is real, not padding.
At the end of W6
A verified cycle in production, then monitoring sits with Agent Care.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Construction AI agent

Build an AP and payroll agent around the signatory who signs.

Show us your timesheets, your determinations and who signs each instrument. Name the three signatures first — the Statement of Compliance, the waiver and the payment schedule — then we map what the agent computes and what it never touches.

Typical delivery
4–6 weeks
Deployment
Your payroll stack
Signatory signs
Compute only
After launch
Agent Care available

Nestack Agents · AP, payroll and back office

AGT-CN-08 · Agent Care available after launch

Build an AP and payroll agent

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

About you
Your payment runs

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