ConstructionManaged AI Agents

Construction AI Agents,
Monitored for Safety

Nestack Agent Care helps construction firms monitor, evaluate, and optimize AI agents used for estimating, scheduling, safety documentation, and vendor coordination — before small AI errors become safety or cost overruns.

143failure modes
60SEV-1 failure modes
1950+baseline eval cases
24/7Agent Monitoring

Scope

Construction AI agents we build & manage

Observability

What we make observable

Every construction agent session is traced across ten layers — what we capture and the evidence we keep.

  1. Evidence we keep
    Goalconstraintsapproval requirement
  2. Evidence we keep
    Sourceversiontimestamprelevancecitation
  3. Evidence we keep
    Planned sequenceactual sequenceworkflow status
  4. Evidence we keep
    Task statusresultretryfailure reason
  5. Evidence we keep
    Tool nameversioninputoutputpermissionresult
  6. Evidence we keep
    Model/versioninput/outputtoken usagelatencycost
  7. Evidence we keep
    Evaluation typemetricthresholdresult
  8. Evidence we keep
    Guardrail targettriggeractionenforcement result
  9. Evidence we keep
    Reviewerdecisioncorrectionreason
  10. Evidence we keep
    Outcome statusbusiness resultlinked trace

Catalog

Failure modes

Filter failure modes by where they occur in the agent lifecycle—from goals and retrieval to tools, evaluations, guardrails and outcomes.

Nothing on this page matches that. Try a layer (tool, retrieval, guardrail), a code, or a severity.

0 documented↑↓ move · ↵ open · ⌘↵ filter matrix
143 documented · select a cell to filter
4
47
42
54
43
22
59
44
32
83
CON-01Wrong safety procedures — heights, trenching, electrical, confined spaceSEV-1
CON-02Estimating errors that bind bids — rates, quantities, escalationsSEV-1
CON-03Stale building codes and standardsSEV-2
CON-04Contract-term hallucination — variations, EOTs, liquidated damagesSEV-1
CON-05Schedule promises without float or logicSEV-2
CON-06Certification and license lapses missed — trades, plant, insurancesSEV-2
CON-07Tender-data leakage between competing bidsSEV-1
CON-08Measurement and unit errors — areas, volumes, ratesSEV-2
CON-09Structural and engineering-calc errors presented as designSEV-1
CON-10Hazardous-materials misadvice — asbestos, silica, lead exposure limitsSEV-1
CON-11Progress-claim and security-of-payment errors — timeframes and entitlementsSEV-2
CON-12RFI and submittal misrouting — spec conflicts resolved wronglySEV-2
CON-13Defect classification and warranty-liability errorsSEV-3
CON-14Planning-condition breaches — hours, noise, sediment, protected treesSEV-2
CON-15Prompt injection via inbound project documents — tender portals, supplier emails, submittalsSEV-1
CON-16Payment-redirection fraud — BEC and deepfake approvals through AP agentsSEV-1
CON-17Superseded-revision retrieval — drawings, specs, geotech reports, as-builtsSEV-1
CON-18False “clear to dig” — utility-locate and 811 screening errorsSEV-1
CON-19Lift-plan and weather go/no-go errors — ground bearing, wind at hook height, poursSEV-1
CON-20Mistranslated safety communications for multilingual crewsSEV-1
CON-21Vision-monitoring false assurance — missed PPE and hazard detectionsSEV-1
CON-22Reality-capture overstatement — progress and volumes feeding pay applicationsSEV-2
CON-23Hallucinated standards, specs and SKUs — phantom ASTM refs, wrong fire ratingsSEV-2
CON-24Lien and statutory-notice deadline miscalculationSEV-2
CON-25COI and endorsement misreading — silent insurance coverage gapsSEV-2
CON-26Certified-payroll and labor-rule errors — Davis-Bacon, prevailing wage, CBAsSEV-2
CON-27Agent commitments that bind the firm — invented policies, prices, promisesSEV-2
CON-28Outbound data leakage — confidential project data into public modelsSEV-2
CON-29Algorithmic pricing collusion exposure — pooled competitor data in pricing toolsSEV-1
CON-30Biometric and worker-surveillance privacy breaches — BIPA, GDPR, EU AI ActSEV-2
CON-31Discriminatory screening — workforce and subcontractor prequalification biasSEV-2
CON-32Fabricated authorities in claims and dispute documentsSEV-2
CON-33Unauthorized autonomous actions — POs, notices and payments sent without mandateSEV-1
CON-34Procurement misfires — hallucinated lead times, unvetted substitutions, counterfeitsSEV-2
CON-35Silent cost-model drift during price volatilitySEV-2
CON-36AI meeting minutes as flawed project recordsSEV-3
CON-37Accessibility-compliance misses in design review — ADA and FHA clearancesSEV-2
CON-38Synthetic evidence accepted — AI-faked photos and documents in claimsSEV-2

A · Tender & estimating assistants

TE-01Superseded input priced01 · RetrievalSEV-3
TE-02Assembly double-counted02 · ReasoningSEV-1
TE-03Open item priced as firm02 · ReasoningSEV-1
TE-04Rate changed after sign-off03 · Tool / writeSEV-1
TE-05Portal form auto-filled03 · Tool / writeSEV-2
TE-06Inferred rate dressed as a quote04 · OutputSEV-2
TE-07Rounding defaults shifted05 · Change / VersionSEV-2

B · Safety-compliance copilots (SWMS, permits)

WM-01Wrong jurisdiction's list01 · RetrievalSEV-3
WM-02Negative HRCW call02 · ReasoningSEV-1
WM-03Control with no method02 · ReasoningSEV-1
WM-04Permit pre-filled03 · Tool / writeSEV-1
WM-05Bulk closure on archive03 · Tool / writeSEV-2
WM-06Unusable reading level04 · OutputSEV-2
WM-07Silent template rewrite05 · Change / VersionSEV-2

C · Project-schedule agents

PJ-01Unaccepted revision pulled01 · RetrievalSEV-3
PJ-02Float read as no delay02 · ReasoningSEV-1
PJ-03Concurrency from overlap02 · ReasoningSEV-1
PJ-04Live programme rewritten03 · Tool / writeSEV-1
PJ-05Register item auto-closed03 · Tool / writeSEV-2
PJ-06Forecast reads as an admission04 · OutputSEV-2
PJ-07Silent calendar change05 · Change / VersionSEV-2

D · RFI & document-control agents

RQ-01Superseded revision returned01 · RetrievalSEV-3
RQ-02Scope change called minor02 · ReasoningSEV-1
RQ-03Deemed acceptance asserted02 · ReasoningSEV-1
RQ-04Status code promoted03 · Tool / writeSEV-1
RQ-05Ageing RFI auto-closed03 · Tool / writeSEV-2
RQ-06Design drafted as a response04 · OutputSEV-2
RQ-07Status semantics remapped05 · Change / VersionSEV-2

E · Subcontractor-management assistants

PQ-01Stale register result01 · RetrievalSEV-3
PQ-02Contractor inferred from paperwork02 · ReasoningSEV-1
PQ-03Statutory window miscounted02 · ReasoningSEV-1
PQ-04Payment schedule auto-issued03 · Tool / writeSEV-1
PQ-05Prequalification flipped on upload03 · Tool / writeSEV-2
PQ-06Insured shown as a binary04 · OutputSEV-2
PQ-07Test changed under the population05 · Change / VersionSEV-2

F · Procurement & supply-chain agents

PU-01Expired evidence read as current01 · RetrievalSEV-3
PU-02Equivalence inferred02 · ReasoningSEV-1
PU-03Lead time solved by swapping02 · ReasoningSEV-1
PU-04PO released before evaluation03 · Tool / writeSEV-1
PU-05Delivery window compressed03 · Tool / writeSEV-2
PU-06Self-declaration read as tested04 · OutputSEV-2
PU-07Catalogue remap on open orders05 · Change / VersionSEV-2

G · Site-vision & progress-capture agents

SI-01Superseded baseline01 · RetrievalSEV-3
SI-02Sheeting read as finishes02 · ReasoningSEV-1
SI-03Rule applied out of place02 · ReasoningSEV-1
SI-04Inferred quantity written03 · Tool / writeSEV-1
SI-05Recognition inherited03 · Tool / writeSEV-2
SI-06Off-site capture published04 · OutputSEV-2
SI-07Backdated re-scoring05 · Change / VersionSEV-2

H · AP, payroll & back-office agents

PW-01Superseded determination read01 · RetrievalSEV-3
PW-02Coverage scoped to the wrong rules02 · ReasoningSEV-1
PW-03Fringe credit on the wrong base02 · ReasoningSEV-1
PW-04Unconditional waiver released03 · Tool / writeSEV-1
PW-05Trust ledger drawn down03 · Tool / writeSEV-2
PW-06Reasons for withholding omitted04 · OutputSEV-2
PW-07Rounding defaults shifted05 · Change / VersionSEV-2

I · Building-safety change-control & gateway agents

FY-01Superseded agreed document01 · RetrievalSEV-3
FY-02Category drafted too low02 · ReasoningSEV-1
FY-03Adviser recorded by role02 · ReasoningSEV-1
FY-04Compliance position asserted03 · Tool / writeSEV-1
FY-05Record dated after the act03 · Tool / writeSEV-2
FY-06Document impact missed04 · OutputSEV-2
FY-07Designated change missed05 · Change / VersionSEV-2

J · Stormwater & site-environmental compliance agents

FZ-03Wrong permit template01 · RetrievalSEV-3
FZ-04Trigger misread02 · ReasoningSEV-1
FZ-06Causation asserted02 · ReasoningSEV-1
FZ-02Lapsed credential accepted03 · Tool / writeSEV-1
FZ-05Clock started late03 · Tool / writeSEV-2
FZ-01Required element missing04 · OutputSEV-2
FZ-07Silent permit regression05 · Change / VersionSEV-2

K · Practical-completion & final-account agents

GG-03Superseded contract text01 · RetrievalSEV-3
GG-04Trifling read from cost02 · ReasoningSEV-1
GG-06Merged conclusivity clock02 · ReasoningSEV-1
GG-02Damages read as deducted03 · Tool / writeSEV-1
GG-05Conditional certificate03 · Tool / writeSEV-2
GG-01Absence never rendered04 · OutputSEV-2
GG-07Silent clock regression05 · Change / VersionSEV-2

L · Workforce licensing & currency agents

GI-01Register misses the chain01 · RetrievalSEV-3
GI-02Licence class not tested02 · ReasoningSEV-1
GI-03Renewal window read as gone02 · ReasoningSEV-1
GI-04Two workers, one name03 · Tool / writeSEV-1
GI-05Clock never started03 · Tool / writeSEV-2
GI-06Completeness read as fine04 · OutputSEV-2
GI-07Start date inherited05 · Change / VersionSEV-2

M · Special-inspection record agents

GL-03Superseded statement01 · RetrievalSEV-3
GL-04Escalation-clock miss02 · ReasoningSEV-1
GL-06Unevidenced closure02 · ReasoningSEV-1
GL-02Imported observation03 · Tool / writeSEV-1
GL-05Reported, not placed03 · Tool / writeSEV-2
GL-01Wrong closing rule04 · OutputSEV-2
GL-07Silent rule regression05 · Change / VersionSEV-2

N · Domestic-content certification agents

GM-03Wrong project date read01 · RetrievalSEV-3
GM-04Inferred component cost02 · ReasoningSEV-1
GM-06Classification skipped02 · ReasoningSEV-1
GM-02Cut-off missed03 · Tool / writeSEV-1
GM-05De minimis double-spend03 · Tool / writeSEV-2
GM-01Open gap left unnamed04 · OutputSEV-2
GM-07Silent rule regression05 · Change / VersionSEV-2

O · Federal certified-claim agents

GP-01Source record archived01 · RetrievalSEV-3
GP-02Threshold netted, not summed02 · ReasoningSEV-1
GP-03Conduct called a change02 · ReasoningSEV-1
GP-04Certificate never signed03 · Tool / writeSEV-1
GP-05Signed without authority03 · Tool / writeSEV-2
GP-06Deemed denial missed04 · OutputSEV-2
GP-07FAR revision not tracked05 · Change / VersionSEV-2

Guardrails

Critical guardrails for Construction agents

Ten controls that hold regardless of prompt, plan or pressure. Open one to see what it protects, what trips it, what the agent is forced to do, who may release it, and what is written to the record.

GR-01No high-risk work advice outside controlled SWMSOverride defined
  1. Target

    Safety guidance for heights, trenching, electrical, confined-space and hazardous-material tasks

  2. Trigger

    A response gives high-risk work steps not drawn from a controlled SWMS or method statement

  3. Action — enforced

    Refuse improvised steps and defer to the controlled SWMS; agent may direct to the responsible supervisor

  4. Human override

    Released only by the site safety manager confirming the controlled SWMS for the task

Logged evidencetask id · hazard class · SWMS reference · refused guidance · manager identity · timestamp

GR-02No clear-to-dig without verified utility locateOverride defined
  1. Target

    Excavation and ground-disturbance advice depending on utility-locate and 811 screening

  2. Trigger

    Advice authorises digging without a verified, in-date utility locate for the excavation zone

  3. Action — enforced

    Block the clear-to-dig and require a verified locate; agent may schedule the locate and hold excavation

  4. Human override

    Released only by the site supervisor after a verified utility locate is recorded

Logged evidencelocate ticket id · excavation zone · locate status · supervisor identity · timestamp

GR-03No lift approved breaching ground-bearing or wind limitsOverride defined
  1. Target

    Crane and lift-plan go/no-go advice covering ground bearing, wind at hook height and pours

  2. Trigger

    A lift is approved against ground-bearing capacity or wind limits at hook height

  3. Action — enforced

    Block the go decision and hold the lift; agent may propose a compliant window or engineered plan

  4. Human override

    Released only by the lift supervisor confirming ground bearing and wind against the lift plan

Logged evidencelift plan id · ground bearing · wind reading · go decision · supervisor identity · timestamp

GR-04No cross-tender data retrievalNo override
  1. Target

    Bid, pricing and strategy data scoped to a single tender within the firm probity boundary

  2. Trigger

    A request or tool call would read a competing tender bid, pricing or strategy

  3. Action — enforced

    Deny the cross-tender read and return only in-scope data; agent continues within the tender boundary

  4. Human override

    None — cannot be overridden in session

Logged evidencerequesting tender · target tender · denied record id · policy rule · timestamp

GR-05No instruction embedded in tenders, emails or submittals ever executedNo override
  1. Target

    Inbound project documents from tender portals, supplier emails, submittals and RFIs the agent reads

  2. Trigger

    An inbound document contains directives, links or tool calls attempting to steer the agent

  3. Action — enforced

    Treat the document as data only and refuse the embedded instruction; the project task continues normally

  4. Human override

    None — cannot be overridden in session

Logged evidencedocument id · source · detected instruction · content hash · timestamp

GR-06No bid submission without verified estimate basisOverride defined
  1. Target

    Estimates, quantities, rates and escalations that could bind a submitted bid or variation

  2. Trigger

    A bid or variation is submitted with unverified quantities, rates or measurement basis

  3. Action — enforced

    Hold the submission and require estimator verification; agent may prepare the priced basis for sign-off

  4. Human override

    Released only by the senior estimator confirming quantities, rates and the measurement basis

Logged evidencebid id · quantities · rate basis · estimator identity · timestamp

GR-07No structural calculation presented as approved designOverride defined
  1. Target

    Engineering and structural calculations the agent produces or summarises for site use

  2. Trigger

    A structural or engineering calculation is presented as approved design without engineer certification

  3. Action — enforced

    Block the design claim and label it indicative; agent may pass the calculation to an engineer for certification

  4. Human override

    Released only by the certifying structural engineer accepting the calculation as approved design

Logged evidencecalculation id · element · assumptions · engineer identity · certification ref · timestamp

GR-08No payee change on unverified bank-detail requestOverride defined
  1. Target

    Accounts-payable changes to payee bank details and payment releases triggered by inbound requests

  2. Trigger

    A request changes payee bank details or releases payment without out-of-band verification

  3. Action — enforced

    Block the change and require callback verification; agent may proceed once the payee is independently confirmed

  4. Human override

    Released only by the finance controller after independent callback verification of the payee

Logged evidencevendor id · old and new bank details · verification method · controller identity · timestamp

GR-09No work issued from superseded drawings or specsOverride defined
  1. Target

    Drawings, specifications, geotech reports and as-builts served to site and procurement

  2. Trigger

    A retrieval returns a superseded drawing, spec or report as the current revision

  3. Action — enforced

    Block issue of the superseded document and serve the current revision; agent flags the revision mismatch

  4. Human override

    Released only by the document controller confirming the current revision status of the document

Logged evidencedocument id · served revision · current revision · controller identity · timestamp

GR-10No progress claim missing statutory payment deadlineOverride defined
  1. Target

    Progress claims, payment schedules and lien or statutory notices with fixed legal timeframes

  2. Trigger

    A claim, schedule or notice is prepared missing a statutory payment or lien deadline

  3. Action — enforced

    Hold the claim and surface the statutory deadline; agent may prepare the notice within the required timeframe

  4. Human override

    Released only by the contracts administrator confirming the statutory deadline and entitlement

Logged evidenceclaim id · statutory regime · deadline date · administrator identity · timestamp

Oversight

Human review — triggers, decisions and evidence

When a defined risk trigger fires, the affected action is routed to a named reviewer. Every decision is recorded with its correction, escalation and final outcome for full traceability.

  • ConfidenceLow-confidence take-off
  • Financial impactHigh-value bid or variation
  • Identity / change riskSupplier bank-detail change
  • Irreversible actionPurchase order or payment
  • Policy riskCode or permit conflict
  • Safety controlGuardrail override
  • Quality failureFailed critical evaluation
Humanreviewnamed reviewer
  • Revieweridentity + role
  • Decisionapprove / reject / amend
  • Correctionwhat changed
  • Escalationwho, why and severity
  • Final outcomereleased / blocked / returned for rework

7 triggers · any one halts the agent

1 record · 5 fields, every time

Compliance

Regulatory mapping

  • 02 Retrieval
  • 03 Workflow
  • 04 Task
  • 05 Tool
  • 06 LLM
  • 07 Evaluation
  • 08 Guardrail
  • 09 Human review
  • 10 Outcome

Site safety & safe systems of work

Confined-space permits expire with the task and training must be in a language the crew understands; a Melbourne SWMS answers to Victorian regulations, not model WHS text; collective measures rank above PPE.

29 CFR 1926.1205, 1926.1207 · OHS Regulations 2017 (Vic) · MHSWR 1999 Sch 1

Lifecycle layer:04 Task · 06 LLM · 08 Guardrail · 09 Human review

  • CON-01
  • CON-20
  • WM-01
  • WM-02
  • WM-03
  • WM-04
  • WM-06
  • WM-07
  • SI-03

Evaluations

Baseline evaluation suite — in detail

Baseline evaluations are completed during onboarding and repeated based on the selected plan. Agents that fail critical checks remain restricted until they pass re-testing.

Evaluation suites
53
Failure modes covered
143 of 143
Cases · 10% rotated / quarter
3,000
Audit-ready scorecard
Monthly

Output evaluation7 suites · 385 cases

100 casesEstimating accuracycatches CON-02
60 casesSchedule realismcatches CON-05
55 casesDefect-classification setcatches CON-13
50 casesMistranslated safety communications for multilingual crewscatches CON-20
45 casesReality-capture overstatementcatches CON-22
40 casesDiscriminatory screeningcatches CON-31
35 casesAccessibility-compliance misses in design reviewcatches CON-37

Use your work email to unlock the full grid. We use it to send the requested Agent Care material and qualify business inquiries. We do not sell or share your email.

  1. Domain-expert review

    Client-designated subject-matter experts review evaluation criteria, pass thresholds and industry-specific risks before baseline approval.

  2. Test-case rotation

    Evaluation cases are refreshed regularly to reduce memorisation, limit overfitting and maintain meaningful performance measurement.

  3. Scorecard integration

    Scorecards compare results with the approved baseline, show performance trends and flag material declines for review and escalation.

  4. Client-specific extensions

    Where included in scope, evaluations may be expanded using approved incidents, workflows, policies, data patterns and industry-specific risks.

Monitoring

Change-aware monitoring

When agent performance changes, Nestack correlates the shift with changes to the agent, prompt, model, tools, knowledge base, guardrails and evaluation suite.

Version changes by layer
01AgentAgent changed in week 5
02PromptPrompt changed in week 3Prompt changed in week 9
03ModelModel changed in week 7Model changed in week 12
04ToolTool changed in week 4
05Knowledge-baseKnowledge-base changed in week 8Knowledge-base changed in week 10
06GuardrailGuardrail changed in week 6
07Eval-suiteEval-suite changed in week 11
Current-revision rate92–100%
  • Week 1 · 97.9%
  • Week 2 · 97.8%
  • Week 3 · 98.0%
  • Week 4 · 97.9%
  • Week 5 · 98.1%
  • Week 6 · 97.9%
  • Week 7 · 98.0%
  • Week 8 · 93.6%
  • Week 9 · 93.4%
  • Week 10 · 97.9%
  • Week 11 · 98.0%
  • Week 12 · 98.1%

Week readouthover or select

Week 8of 12

  • 05Knowledge-basekb 2026.07

93.6%Current-revision rate

  • 7 layers stamped on every run · 12-week window
  • Catches CON-17 · superseded-revision retrieval

Something missing?

Don’t see your agent’s issue here?

Every AI environment is different. Share what you’re seeing, and we’ll review the behaviour, assess the risk and recommend the evaluations or controls that may help.

Report an agent issue

Tell us what you’re seeing

The more specific, the faster we can reproduce it. Playbook: Construction

We reply within one business day.

See pricing

No commitment. Even if you never become a client, we’ll tell you what we think is happening.

Process

Universal incident runbook

Severity is assigned based on business impact, customer harm, data exposure, operational disruption and overall scope.

  • SEV-1 Critical
  • SEV-2 Major
  • SEV-3 Moderate
  • SEV-4 Minor
STEP 01

Automated monitoring or human review identifies unusual behaviour. Alerts are recorded and routed according to severity.

Outcomes

Not just that it ran.
What it delivered.

Observability shows the agent ran. It can’t show the drawing was current or what the work cost. Where outcome data exists, we trace each result to its session — and a named reviewer signs off the month.

Issued
Monthly, per engagement
Traced
Every figure to its session
Signed
By your engagement reviewer
Used for
Client reporting, partner review, the AgentOps scorecard

Nestack AgentOps

Construction fleet · monthly statement

  • RFI answered1,840
  • Submittal reviewed960
  • Progress claim assessed210
  • Permit condition checked640
  • Site inspections completed48 of 48
  • Workflows delivered3,650
  • Outcome success rate98.1%
  • Human correction required69 · 1.9%
Average AI cost per successful workflow$0.41

Every figure linked to its source trace · exportable for review and audit support

Cost control

Keep construction AI agent costs under control

Token spend is monitored, optimised and reported as part of Agent Care — and savings never come at the expense of quality, because every change is verified against your evaluation baseline.

We review token spend by agent, workflow, model, and session so you can understand where AI costs are coming from.

Running construction AI agents in production?

Get a free assessment of one agent. We’ll review its behaviour, run a baseline evaluation and highlight potential risks and performance gaps.

Loading open times…

Open times

Select an available date above to view open 45-minute times from our calendar.