Main site

Construction AI agent · Project schedule

Project-Schedule & Delay-Notice AI Agent

A time bar usually bars: a notice missed by a day loses the entitlement on merits nobody reaches. Where a court may now relieve one, the evidence decides, and the agent keeps it.

Typical delivery
4–6 weeks
Deployment
Your stack
Planner review
Drafts only
After launch
Agent Care

What this agent does

Computes the deadline; it does not serve the notice

In

  1. 01

    Compute the notice-deadline register from the clocks and triggers your executed contract sets.

  2. 02

    Diff each programme revision against the last — activity, duration, tie and calendar.

Reason

  1. 03

    Archive every revision immutably, with its author, timestamp, data date and provenance.

  2. 04

    Compute time elapsed on each open item and escalate long before the window closes.

  3. 05

    Diff field progress against the model and flag as-built dates no record supports.

Decide

  1. 06

    Diff the model against the accepted programme and flag drift instead of closing it.

  2. 07

    Route every register item and every draft to the planner who holds it.

Out

  1. 08

    Archive the records read, the draft, the amendments and the decision against the event.

  2. 09

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

Product statement

The agent computes deadlines and drafts; the contracts administrator serves the notice — and a date that has passed cannot be un-passed by anyone.

Example workflow

One delay event, record to draft

  1. 1Delay event recordedSite diary, progress update, instruction or correspondence
  2. 2Clocks resolvedThe notice periods, triggers and escalation steps configured from the contract as executed
  3. 3Register item computedDeadline, time elapsed and confidence
  4. 4Controls appliedConfigured-clock checks, evidence checks, escalation timers and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is served at any of them — the agent is computing deadlines, and the planner's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the planner who holds it.

Low confidence

Adds a contracts read first.

Planner review

The item is held with its deadline, the records behind it and the confidence.

Acknowledge · Amend draft · Send to contracts

Acknowledged — back with contracts

  1. 6Schedule systems updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedEscalation lead time, drafts amended, notices served and events that ran past the window

Amendments

Every amendment made in review is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Serving or transmitting any contractual notice.
    • Approving, rejecting or accepting a programme submission.
    • Changing the baseline, data date, calendars or logic.
    • Selecting the delay-analysis method.

  • Automation boundary

    Agent acts unaided

    • ✓Compute the notice-deadline register with escalation timers for each named owner.
    • ✓Diff two programme revisions, activity by activity.
    • ✓Reconcile field progress against the model.
    • ✓Archive every revision with its author, timestamp and provenance.

    Any write happens inside the boundaries agreed at implementation, and never into the live programme.

    • Characterising delay as excusable, compensable or concurrent.
    • Certifying, waiving or forecasting liquidated damages.
    • Declaring time at large, or a claim time-barred.
  • Releasing a schedule or forecast outside the project.

Example output

One register item, annotated

Everything the agent computes is attached to the records it was read from.

Register item · single delay event

Illustrative example
Package
Piling subcontract
Delay event
Access to the piling zone withheld by the principal
Window elapsed
52% elapsed
Item state
Escalated — draft with the planner
Confidence
93%
Clock source
The contract as executed

As received

Taken from the site records and the clocks configured for this contract — nothing on this side is written by the agent.

What was read

  • Site diary entry
  • Accepted programme revision
  • Correspondence log

Why this deadline

The period comes from the contract as executed and amended.

What the score decides

Below the configured threshold the item picks up a contracts read before the planner sees it.

Action

AcknowledgeAmend draftSend to contracts

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 delay event

From records and the contract

03Deadline

Apply the contract as executed

Standard forms are almost always amended, so the agent works from the clocks configured for your contract, not the printed form.

01Approved path

See the bar before it closes

Routine register items and revision diffs arrive already computed.

02Human review

Send review to the tight clocks

Short windows and low-confidence items are marked, so the planner's read starts where the clock is tightest.

04

Build an evidence trail

Every revision keeps its author, its timestamp and the diff that produced it.

Integrations

Typical integrations

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

  • CPM scheduling engines

    • Oracle Primavera P6
    • Asta Powerproject
    • Microsoft Project
    • Deltek Open Plan

  • Schedule analytics

    • Deltek Acumen Fuse
    • SmartPM
    • Revision exports and progress updates

  • Field progress capture

    • Procore
    • Autodesk Construction Cloud
    • Site diaries and daily records

Agent

Project schedule & delay notices

  • Reads the records
  • Computes the deadlines
  • Holds for the planner

  • Contract and document backbone

    • Oracle Aconex
    • Trimble Viewpoint
    • Notice and contract 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 notice

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

L6Rollback / safe modePull the agent back to register-only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, calendar-default and contract-configuration changes.Track
L4TraceabilityRecord the revision read, the clock applied, the reviewer and the decision behind each item.Record
L3Planner reviewHold items for the named planner; it governs release, not whether the notice later served is right.Gate
L2Policy guardrailsTest items against the configured clocks and the write boundary; a failure returns the item.Restrict
L1Confidence thresholdsRoute low-confidence items to a contracts read before the planner sees them.Require review
Model coreItem produced — deadline, time elapsed, records gathered and confidence
L1 – L2
Test whether an item may stand
L3
Puts the decision in the planner's hands
L4 – L5
Hold the revision history the claim rests on
L6
Stops drafting when evaluation signals slip

How Nestack evaluates it

Evaluate the notice workflow — not only the finished programme.

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

Surface — the deadline the project acts on

  1. E1Final-output evaluationWas the deadline computed from the clock this contract actually sets?
  2. E2Step-level evaluationDid the agent read the right revision, records and contract configuration?
  3. E3Tool evaluationDid it read and write the correct project and the correct register?
  4. E4Confidence calibrationDo low-confidence items actually attract more planner amendments?
  5. E5Slice evaluationHow does performance change across specific contract forms?
  6. E6Business outcomeHow many events needed an amendment, or ran past the escalation window?

Floor — the entitlement the contractor keeps

Failure modes

Where each failure originates in the agent

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

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • PJ-01

      Unaccepted revision pulled

      Analysis runs against a programme that was never the datum.

    Stage gathers

    Revisions, site records, progress updates and configured clocks

  2. 02 · Reasoning

    2 modes

    • PJ-02

      Float read as no delay

      Float consumed by an employer risk event is reported as no impact.

    • PJ-03

      Concurrency from overlap

      Calendar overlap is read as concurrent delay to completion.

    Stage proposes

    Deadline, time elapsed and confidence

  3. 03 · Tool / write

    2 modes

    • PJ-04

      Live programme rewritten

      A revised data date and re-levelled resources erase the as-planned logic.

    • PJ-05

      Register item auto-closed

      An item closes as no entitlement and the escalation clock stops.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • PJ-06

      Forecast reads as an admission

      Client-facing wording concedes contractor culpability.

    Stage returns

    The register the planner reads before serving

  5. 05 · Change / Version

    1 mode

    • PJ-07

      Silent calendar change

      A tool upgrade shifts retained logic, so the longest path moves.

    Stage tracks

    Model, prompt, calendar defaults and contract config

  • Sev-1 · a held act performed by the agent
  • Sev-2 · a wrong deadline reaches the planner
  • Sev-3 · records degrade, item routes to review
See the complete Construction failure-mode catalogue

Affected slices

A steady average can rest on one contract form

The all-event baseline is the number a monthly report shows; underneath it a small set of contract cohorts absorbs most of the late escalations. Nestack reports the rate by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
FIDIC and NEC EPC packages5.9%3.7×Review
AS 4000 head contracts4.0%2.5×Review
JCT and A201 traditional build2.4%1.5×Watch
Framework programmes0.8%0.5×Normal

Bar: late-escalation-rate lift vs. an all-event baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

Nothing closes until the suite grows

The loop shuts when the miss is a case in the suite, not when someone has explained it.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Late-escalation rate rises in a contract-form slice.

  2. 02

    Diagnose

    Not isolated until the archive, the records and the register agree on one cause.

  3. 03

    Improve

    Whatever changes ships against a version, with the notices that prompted it attached.

  4. 04

    Verify

    Nothing ships until the affected cases pass a second time.

  5. 05

    Learn

    The suite grows by one case; so does the notice playbook.

  6. Learn → Detect

    The return edge. Detection next time runs against a longer suite.

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

Workstream

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

  1. 01Programme workflow discovery and boundary work.Week 1
  2. 02Schedule and field-record assessment.Week 1, Week 2
  3. 03Contract-clock, calendar, float and notice-rule mapping.Week 1, Week 2
  4. 04Programme ingestion and normalisation.Week 2
  5. 05Deadline computation and revision-diff logic.Week 2, Week 3
  6. 06Confidence scoring and register-item routing.Week 3
  7. 07Planner review workflow.Week 3, Week 4
  8. 08Scheduling and document-system integration.Week 5
  9. 09Notice-timing regression cases.Week 4, Week 5
  10. 10Guardrails and service controls.Week 4
  11. 11Revision-archive 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 scheduling systemsAdvancedMultiple projects / forms
Introduced at Pilot
Deadlines from your executed contract✓✓✓
Planner review✓✓✓
Notice-timing baseline✓✓✓
Introduced at Production
Reporting by contract form—✓✓
Review workflow in your systems—✓✓
Approved write-back—✓✓
Client and funder reporting feed—✓✓
Introduced at Advanced
Multi-form portfolios——✓
Multi-stage contract review——✓
High event volume——✓
Enterprise programme controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, event 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 executed amendments

    Contract-clock and trigger mapping

    Week 1

  2. 02Representative programme revisions

    Deadline baseline, revision diffing and progress reconciliation

    Week 2

  3. 03Your notice register and escalation practice

    Escalation-timer mapping and automation-boundary definition

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Scheduling, field and document assessment, then integration setup

    Week 2

  5. 05Delay events you did not notice in time

    Notice-timing cases and the evaluation suite

    Week 4

  6. 06Where a draft notice stops and waits for the planner

    Confidence scoring, item routing, guardrails and approval controls

    Week 3

  7. 07Named planners to review the register

    Planner review 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

Each phase is drawn across the weeks it truly occupies. Week 5 carries evaluation and pilot at once.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Programme workflow discovery, clock mapping and the automation boundary

W2Source integration and the deadline baseline

W3Diff logic, confidence scoring and review controls

W4Evaluation suite, guardrails and failure-mode testing

W5Scheduling-system integration, pilot events and targeted corrections

W6A full monthly programme cycle under the planner, then handover

Reading the band
Each bar covers the weeks its work is named in and no others. The week 5 overlap is real, not padding.
At the end of W6
Once the cycle validates, Agent Care owns the running agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Construction AI agent

Build a project-schedule agent around your notice clocks.

Show us your contracts as executed, your programme revisions and who serves the notice. Once a window closes the entitlement closes with it, so we map the clocks first and name what the agent never touches.

Typical delivery
4–6 weeks
Deployment
Your scheduling stack
Planner review
Drafts only
After launch
Agent Care available

Nestack Agents · Project schedule and delay notices

AGT-CN-03 · Agent Care available after launch

Build a project-schedule agent

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

About you
Your projects

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