Water Discharge & Exceedance Certification AI Agent

What this agent does

Assembles the pack, not the certification

A sample is taken at an outfall, and the custody form, the sampler named on it and the laboratory job come with it.

A result returns, and its QC status, its holding time and the analytical method are carried with the value.

Reason

A value is compared against the permit limit held for that outfall, on the averaging basis that permit sets.

A limit is exceeded, and the pack fixes the outfall, the value and the time the result was first read.

A clock runs from that entered time, with the oral and written windows of 40 CFR 122.41(l)(6) beside it.

Decide

A reportability question stays open on the pack, marked open, until a named person answers it.

A non-compliance narrative and a corrective-action record are drafted from the logs the site kept at the time.

Out

A record is retained under 40 CFR 122.41(j) for the period it sets, naming who sampled and who analysed.

A write runs only inside the approval boundaries agreed at implementation, and never ahead of the signature.

→ Product statement

The agent assembles and compares; the named officer certifies under 40 CFR 122.22(d) and personally carries what a knowing false statement attracts.

Example workflow

One outfall, sample to certification

AgentHuman

1Sample receivedField sampling round, contractor custody form or laboratory portal

2Pack assembledCustody records, laboratory data and QC, permit limits and site logs, each with its source

3Comparison runOutfall values, limit basis, flagged exceedances and confidence

4Controls appliedCustody and QC checks, limit-basis checks, signatory-authorisation checks and confidence threshold

No human action required

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

5DecisionBranches at the confidence threshold

High confidence

Goes to the named officer to certify.

Low confidence

Adds an environmental-lead read first.

Officer certification

The pack is held with its evidence, its open questions and the confidence.

Approve · Amend · Send to second read

Certified — report released▼

6Reporting system updatedOnly where write access and approval policy allow it

7Outcome evaluatedPack completeness, amendment rate, notification timeliness and inspection findings

Amendments

A corrected report goes out, and the late notification is itself the violation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Signing the certification statement on the report.

Deciding whether an event is reportable.

Making the oral or written notification to the regulator.

Judging whether an upset or bypass defence holds.

Automation boundaryAgent acts unaided

✓Assemble the pack from custody, laboratory and permit records.

✓Compare each outfall value against the limit on file.

✓Run the notification clock forward from the entered time.

✓Name what is unresolved and hold the pack for the certifying officer to read.

Writes stay inside the boundaries agreed at implementation, and never ahead of the certification.

Authorising a signatory, or filing that authorisation.

Accepting a laboratory result that failed its QC.

Approving a corrective action or calling it complete.

Changes to limit rules, clocks or approval thresholds.

Example output

One outfall value, annotated

Everything the agent assembles is attached to the record it was drawn from.

Certification pack output · single outfallIllustrative example

Outfall

Drafted line

Clock cited

Source of time

Confidence

Clock

Treatment-circuit outfall

The monthly value sits above the limit written for this outfall

40 CFR 122.41(l)(6)

Officer-entered

88%

Runs from the entered time

As receivedTaken from the custody, laboratory and permit records — nothing on this side is written by the agent.

Evidence used Chain-of-custody form Laboratory QC batch Permit limit page

Why this basisIt is the averaging basis the permit sets, and a person still decides what follows.

ActionApproveAmendSend to second read

What the score decidesBelow the configured threshold the pack picks up an environmental-lead read.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04

Every outfallFrom the sample record

03Assembly

Take the part tenement tracking

The tenement agent tracks what a tenement obliges you to do; this one assembles the report an officer certifies.

01Approved path

A named officer certifies

Routine monthly values arrive already compared, with their evidence attached.

02Human review

Send review to the exceedances

A flagged value, an unresolved QC batch or a signatory that no longer matches is marked before the officer reads.

04Build an evidence trail

The sample, the limit it was measured against and the officer who certified stay on the report.

Integrations

Typical integrations

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

Laboratory dataLIMS · EDD deliverables
Contract laboratory portals

Field and samplingCustody forms · field meters
Site rain and flow gauges

Permit and limitsPermit register · limit tables
Outfall and monitoring plan

Agent

Discharge certification

Reads the sample
Compares to the limit
Holds for certification

Reporting and recordsNetDMR · state e-reporting
Records archive · retention

Observability & evaluationOpenTelemetry · 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 report

The six nest tightly, outer to inner. What escapes all six is listed in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modeHold assembly at draft-only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, limit-rule and permit-configuration changes.Track

L4Pack traceabilityRecord the sources, the comparison, the amendments and what stayed unresolved.Record

L3Officer certificationHold packs for the named officer; the gate governs release, not whether the certified value is right.Gate

L2Policy guardrailsTest packs against custody, QC, limit-basis and signatory rules; a failure returns the pack.Restrict

L1Confidence thresholdsRoute low-confidence packs to an environmental-lead read before the officer sees them.Require review

Model corePack assembled — outfall values, limit basis, flagged exceedances and confidence

L1 – L2Test whether a value may stand

L3Puts the certification in the officer's hands

L4 – L5Keep the sample and the limit behind it

L6Holds the report at uncertified when signals degrade

How Nestack evaluates it

Evaluate the assembly workflow — not only the finished report.

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

Surface — the report the regulator reads

Depth of coverage ▼

E1Final-output evaluationDid each value in the pack match the record it came from?

E2Step-level evaluationDid the agent use the right outfall, the right limit and the right basis?

E3Tool evaluationDid it read and write the correct outfall and the correct field?

E4Confidence calibrationDo low-confidence packs actually attract more officer amendments?

E5Slice evaluationHow does performance change across specific outfall types?

E6Business outcomeHow many packs reached the officer unresolved, and how many reports were corrected after?

Floor — the outcome the permittee answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed at the stage where it originates.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

DM-03

Superseded limit value

The limit came from a record the modification replaced.

Stage gathersCustody records, laboratory data and permit limits

02 · Reasoning2 modes

DM-04

Awareness timed to the inbox

The clock is dated from when the report was opened.

DM-06

Averaging basis misapplied

A monthly average is run against a daily limit.

Stage proposesOutfall values, the limit basis and confidence

03 · Tool / write2 modes

DM-02

Reportability closed by silence

An open question leaves the pack as though settled.

DM-05

Sample type not checked

A grab stands where the permit sets a composite.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

DM-01

Signatory not on file

The named signer does not match the authorisation filed.

Stage returnsThe pack the officer certifies and the regulator reads

05 · Change / Version1 mode

DM-07

Silent limit-rule regression

A model or rule change loosens what is flagged.

Stage tracksModel, prompt, limit rules and permit config

Sev-1 · certified outside the boundary Sev-2 · a wrong value reaches the report Sev-3 · lab data degrades, pack holds

See the complete Mining failure-mode catalogue →

Affected slices

One outfall can carry most of the exceptions

An outfall-level certification-readiness figure can look settled while a few outfalls absorb most of the unresolved items. Nestack reports the unresolved-item rate by outfall, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Storm-driven outfalls	9.1%	3.6×	Review
Pit-dewatering discharges	6.9%	2.8×	Review
Treatment-circuit outfalls	4.7%	1.9×	Watch
Steady-state outfalls	2.3%	0.9×	Normal

Bar: unresolved-item lift vs. the steady-state-outfall baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

Nothing closes until a test exists

A cycle closes when the late notification is a regression case. That suite is what the next report certified is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Late notifications rise on one outfall.

02Diagnose

The exceedance known on site before the watched clock started is read back until one cause remains.

03Improve

The fix carries a number, and the outfalls that forced it travel with it.

04Verify

Nothing goes out until every touched outfall case has passed again.

05Learn

It is kept permanently, and the limit rules are amended alongside it.

Learn → DetectThe return edge. Each new cycle opens on a suite that is 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, assembly workflow, evaluation, integration, then production validation and handover.

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

01Discharge workflow discovery and boundary definition.

02Laboratory and custody source assessment.

03Permit limit, averaging-basis and clock rule mapping.

04Sample and result ingestion, normalised.

05Outfall comparison and source binding.

06Confidence scoring and exception routing.

07Officer certification workflow.

08Laboratory and reporting integration.

09Limit and exceedance cases.

10Guardrails and certification controls.

11Sample-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 tier PilotOne outfall, one permit ProductionProduction laboratory systems AdvancedMultiple permits / sites

Introduced at Pilot

Assembly to your permit record✓✓✓

Officer review gate✓✓✓

Chain-of-custody baseline✓✓✓

Introduced at Production

Reporting by outfall—✓✓

Review workflow in your systems—✓✓

Approved write-back—✓✓

Laboratory-system integration—✓✓

Introduced at Advanced

Multi-permit limit rules——✓

Multi-stage officer approvals——✓

High outfall volume——✓

Multi-permit certification controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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.

You bringWe build with it

01Your outfall list and monitoring plan structure → Sample and result ingestion and evidence mappingWeek 1

02Representative laboratory deliverables → Assembly baseline, limit extraction and source bindingWeek 2

03Your permit limits, clocks and signatory register → Limit, clock and signatory-rule mapping, and the boundaryWeek 1

04Access to relevant APIs, feeds or exports → Laboratory and custody assessment, then integration setupWeek 2

05Reports you would not want re-sampled → Clock cases and the evaluation suiteWeek 4

06What no certification may cover over → Confidence scoring, exception routing, guardrails and approval controlsWeek 3

07Named officers to certify the packs → Officer certification workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Weeks, not slides, set the width of every band on this chart, which is why the fifth carries a pair.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Discharge workflow discovery, limit mapping and the automation boundary W2Laboratory integration and the assembly baseline W3Assembly workflow, confidence logic and certification controls W4Evaluation suite, guardrails and failure-mode testing W5Reporting integration, pilot outfalls and targeted corrections W6One reporting period run under the responsible officer, then Agent Care handover

Reading the bandA bar covers only the weeks its work is named in. The week 5 overlap is two phases at once, not padding.

At the end of W6The final checks clear on live outfalls and monitoring moves to Agent Care.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Mining AI agent

Build a discharge-certification agent around your permit and your signatory.

Show us your outfalls, your laboratory feed and your permit limits. Who carries the report when it is wrong — the officer whose name is on the certification, which is why the pack is built for that signature.

Discuss this agent →

Nestack Agents · Discharge certificationAGT-MN-12 · Agent Care available after launch