Main site

Construction AI agent · Completion

Practical-Completion & Final-Account AI Agent

Four contract forms close a final account differently: seven days under AS 4000 cl.37.4, 28 under JCT, four weeks under NEC4, no clock at all under FIDIC. The agent counts; the certifier signs.

Typical delivery
4–6 weeks
Deployment
Your stack
Certifier signs
Per defect
After launch
Agent Care

What this agent does

Builds the completion record, never the certificate

In

  1. 01

    A defect is raised by direction, instruction or notification — a ranked photograph is an observation, not a Defect.

  2. 02

    A register item carries its dated evidence, its instrument, the trade that owns it and the subcontract behind it.

Reason

  1. 03

    A trifling or not-trifling reading is proposed per item — Mears in England and Wales, AS 4000 cl.1 in Australia.

  2. 04

    A completion test is read from this contract, whether that is AS 4000 cl.1, AIA §9.8.1 or NEC4 cl.11.2(2).

  3. 05

    A window opens on certification, and each form is counted on its own rule rather than on a merged one.

Decide

  1. 06

    A final-account line is matched to its instruction, valuation basis or deduction instrument, or flagged unevidenced.

  2. 07

    A statutory completion instrument is kept in a separate register, never allowed to stand for a contractual one.

Out

  1. 08

    A draft rationale, its evidence and each person who saw it before the certifier did are retained on the item.

  2. 09

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

Product statement

The agent proposes and counts; the named superintendent, architect, contract administrator or engineer certifies — and a certificate wrongly issued has no correcting entry, only an adjudication.

Example workflow

One register item, raised to signature

  1. 1Defect raisedA superintendent direction, a contract-administrator instruction or an NEC4 notification
  2. 2Evidence assembledDated photographs, the instrument that notified it, the owning trade and the subcontract behind it
  3. 3Reading proposedTrifling or not, the reason, the contract test cited and confidence
  4. 4Controls appliedContract-form checks, instrument checks, per-form clock rules and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is certified at any of them — the agent is proposing, and the certifier's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the named certifier to decide.

Low confidence

Picks up a contract-administration read first.

Certifier decision

The item is held with its evidence, the test cited and the confidence.

Accept · Reclassify · Refer to the contract administrator

Accepted — the reading stands

  1. 6Project systems updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedItems reopened, windows closed unworked, unevidenced lines and post-certificate disputes

Reclassifications

Every certifier reclassification is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Certifying practical or substantial completion.
    • Determining that a defect is trifling.
    • Issuing, valuing or agreeing a final account.
    • Deducting liquidated damages or calling on security.

  • Automation boundary

    Agent acts unaided

    • ✓Assemble the defects and outstanding-work register from the record.
    • ✓Bind each item to its instrument, its evidence and its owner.
    • ✓Count each conclusivity clock on that form's own rule.
    • ✓Reconcile each account line to an instrument; flag the ones without.

    Any write stays inside the boundaries agreed at implementation, and never touches a certificate.

    • Issuing a certificate of making good.
    • Serving or declining a notice of dispute.
    • Releasing retention or issuing a payment instrument.
  • Changes to contract-form, clock or threshold rules.

Example output

One register item, annotated

An item is attached to the instrument that made it a Defect, not to the photograph that found it.

Completion output · single register item

Illustrative example
Item
Level 6 lift lobby
Outstanding work
Chipped stone threshold at the lift lobby, the lobby otherwise usable
Window remaining
6 clear days
Instrument
Superintendent direction
Confidence
86%
Contract form
AS 4000-1997 cl.37.4

As received

Taken from the register, the direction and the contract itself — nothing here is decided by the agent.

Evidence on the item

  • Direction reference
  • Dated site photograph
  • Subcontract expiry

Why not trifling

Read against the AS 4000 cl.1 use test, not cost — the superintendent calls it.

What the score decides

Below the configured threshold the item picks up a contract-administration read first.

Action

AcceptReclassifyRefer to the contract administrator

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 register item

From the site record

03Assembly

Read against this contract

Draw on the register, the instruments already issued and the completion test this contract actually sets.

01Approved path

Trifling is a human word

Routine items arrive evidenced, owned and dated before anyone opens the file.

02Human review

Put the read where the argument is

Contested items carry their reason and their clock, so the certifier starts where the money concentrates.

04

Build an evidence trail

The defect, the test it was read against and the certifier who signed stay on the register.

Integrations

Typical integrations

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

  • Project management

    • Procore
    • Autodesk Build
    • Aconex
    • Asite

  • Defects and snagging

    • Snagr
    • Fieldwire
    • PlanRadar
    • Dalux

  • Cost and final account

    • CostX
    • Candy
    • RIB
    • Cubit Estimating

Agent

Practical completion & final account

  • Reads the register
  • Counts the clocks
  • Holds for the certifier

  • Contract administration

    • Contract-form templates
    • Handover
    • O&M registers

  • 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 certificate

Six screens tightening inward. What survives them all is named in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modePull assembly back to proposal-only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, contract-form and clock-rule configuration changes.Track
L4TraceabilityRecord the item, its instrument, the proposed reading and who read it first.Record
L3Certifier decisionHold items for the named certifier; it governs who decides, not whether the reading was right.Gate
L2Policy guardrailsTest each item against this contract's own form and clock rules; a failure returns the item unread.Restrict
L1Confidence thresholdsSend low-confidence items to a contract-administration read before the certifier.Require review
Model coreItem proposed — evidence, instrument, reading, clock and confidence
L1 – L2
Test whether a reading may stand
L3
Puts the call in the certifier's hands
L4 – L5
Keep the defect and the test behind it
L6
Narrows to defects-register assembly when signals degrade

How Nestack evaluates it

Evaluate the whole completion workflow — not only the register at the end.

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

Surface — the record the certifier signs against

  1. E1Final-output evaluationDid each register item carry the instrument the contract requires?
  2. E2Step-level evaluationDid the agent read the right contract form, revision and completion test?
  3. E3Tool evaluationDid it read and write the correct project, item and clock?
  4. E4Confidence calibrationDo low-confidence readings actually attract more reclassifications?
  5. E5Slice evaluationHow does performance change across contract forms and defect types?
  6. E6Business outcomeHow many items were reopened, and how many windows closed unworked?

Floor — the outcome the certifier answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each set where it began and not where it landed.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • GG-03

      Superseded contract text

      The item is read against an unamended standard form.

    Stage gathers

    The register, the instruments issued and the contract

  2. 02 · Reasoning

    2 modes

    • GG-04

      Trifling read from cost

      An item is called trifling because it was cheap to fix.

    • GG-06

      Merged conclusivity clock

      One day count runs across four forms; FIDIC has none.

    Stage proposes

    Evidence, instrument, reading and confidence

  3. 03 · Tool / write

    2 modes

    • GG-02

      Damages read as deducted

      A computed liquidated-damages figure is set off.

    • GG-05

      Conditional certificate

      A hedged certificate is logged as achieved.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • GG-01

      Absence never rendered

      The pack lists what arrived and is silent on what never did.

    Stage returns

    The completion record the certifier signs against

  5. 05 · Change / Version

    1 mode

    • GG-07

      Silent clock regression

      A rule change quietly moves one form's conclusivity window.

    Stage tracks

    Model, prompt, contract-form and clock rules

  • Sev-1 · acts on a certifying decision
  • Sev-2 · a window closes unworked
  • Sev-3 · evidence degrades, item goes to review
See the complete Construction failure-mode catalogue

Affected slices

The number holds until you split it by form

A project-level register-evidence figure can read well while one contract form and one defect type carry most of the unevidenced items. Nestack reports the unevidenced-item rate by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Amended AS 4000 and AS 49027.4%3.6×Review
Latent and non-apparent defects5.6%2.7×Review
NEC4 defects-date projects3.1%1.5×Watch
Unamended standard forms1.3%0.6×Normal

Bar: unevidenced-item-rate lift vs. an unamended-form baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

A cycle is finished when a test exists

A cycle is done when the late dispute notice has become a case the next release must pass. That suite is what the next account closed is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Unevidenced items rise on one contract form.

  2. 02

    Diagnose

    The item the builder called trifling and the certifier did not is read back until the cause is one thing.

  3. 03

    Improve

    Number the change; the certificates it touched are filed under it.

  4. 04

    Verify

    Each touched defect case is run once more, and one red holds it back.

  5. 05

    Learn

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

  6. Learn → Detect

    The return edge. Each account closed leaves the suite harder to pass.

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

Workstream

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

  1. 01Completion workflow discovery and boundary setting.Week 1
  2. 02Register and cost-source assessment.Week 1, Week 2
  3. 03Contract-form, clock and instrument rule mapping.Week 1, Week 2
  4. 04Register ingestion and item normalisation.Week 2
  5. 05Reading logic and instrument binding.Week 2, Week 3
  6. 06Confidence scoring and item routing.Week 3
  7. 07Certifier decision workflow.Week 3, Week 4
  8. 08Cost-system and register integration.Week 5
  9. 09Defect and conclusivity cases.Week 4, Week 5
  10. 10Guardrails and certifying controls.Week 4
  11. 11Register-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 contract formProductionProduction project systemsAdvancedMultiple forms / portfolios
Introduced at Pilot
Assembly to your contract form✓✓✓
Certifier decision✓✓✓
Register-evidence baseline✓✓✓
Introduced at Production
Reporting by contract form—✓✓
Conclusivity clocks in your systems—✓✓
Approved write-back—✓✓
Contract-and-cost integration—✓✓
Introduced at Advanced
Amended and bespoke forms——✓
Multi-stage certifier review——✓
High project volume——✓
Multi-form conclusivity 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 contract forms and their amendments

    Register ingestion and item mapping

    Week 1

  2. 02Representative registers from live projects

    Reading baseline, instrument binding and clock rules

    Week 2

  3. 03Your certifiers, by name and appointment

    Contract-form, clock and instrument rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Register and cost-source assessment, then integration setup

    Week 2

  5. 05Accounts you would not want re-opened

    Window cases and failure-mode testing

    Week 4

  6. 06What no final certificate may absorb

    Confidence scoring, item routing, guardrails and certifying controls

    Week 3

  7. 07Named certifiers to decide each reading

    Certifier 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

What sets a band is the time the phase truly takes, and that is why two of them share week five.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Completion workflow discovery, form mapping and the boundary

W2Register and cost-source integration and the reading baseline

W3Reading workflow, clock logic and certifying controls

W4Evaluation suite, window cases and failure-mode testing

W5Cost-system integration, pilot projects and targeted corrections

W6One completion cycle run under the named certifier, then Agent Care handover

Reading the band
Each bar stops where its named work stops. The week 5 overlap is scheduled, not padding.
At the end of W6
When the final account validates, Agent Care takes the agent on.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Construction AI agent

Build a completion agent around your contract, not around a house default.

Bring your contract forms, the register you actually keep and whatever your legal team amended out of clause 37. The certificate is carried personally by whoever signs it, long after the job closes. Accounts payable owns the payment instruments; this evidences what they pay against.

Typical delivery
4–6 weeks
Deployment
Your project and cost systems
Certifier signs
Per defect
After launch
Agent Care available

Nestack Agents · Practical completion & final account

AGT-CN-11 · Agent Care available after launch

Build a completion agent

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

About you
Your project

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