Nestack Agent Care

Energy AI agent · Drinking water

Drinking Water Compliance AI Agent

Record the sample, the limit it is measured against and the rule version its schedule runs under — then hold the compliance record for the person your state primacy agency names to sign.

4–6 weeksTypical delivery
Your stackDeployment
Version-boundState names it
Agent CareAfter launch

What this agent does

Drafts the record, never certifies it

In
01

A result returns, and 40 CFR 141.31(a) allows the first ten days of the following month.

02

A limit is exceeded, and § 141.31(b) leaves 48 hours to report the failure to the State.

Reason
03

A notification completes, and § 141.31(d) wants the certification within 10 days of it.

04

A lead action level is exceeded, and § 141.31 caps the copy at 24 hours, sooner if practicable.

05

A report is distributed, and § 141.155(c) wants a copy and a certification within 10 days.

Decide
06

A report is drafted, and the revision published 24 May 2024 applies from 2027, not the 2023 proposal.

07

A schedule is set, and § 141.80(a) is effective 30 December 2024 with compliance by 1 November 2027.

Out
08

A PFAS sample is drawn, and the April 2024 rule stands whole; both May 2026 rules are proposals.

09

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

Product statement

The agent assembles and reconciles. Part 141 names no federal signer — the state primacy agency names one, and that person certifies and submits.

Example workflow

One sample, result to release

AgentHuman
1Sample result receivedLaboratory reports, inventory entries, monitoring schedules or notification triggers
2Sample context assembledThe system, the method and holding time, the averaging basis and the rule version in force
3Compliance record draftedThe sample, its result, the gaps and completeness
4Controls appliedClock checks, record-sufficiency checks, version checks and completeness confidence
No human action required

Stages 1 to 4 run unaided, and nothing is certified at any of them — the agent is assembling, and the compliance lane opens at the completeness gate.

5DecisionForks at the completeness gate
Record sufficient

Goes to the signer the primacy agency names.

Anything thin

Adds a water quality manager read first.

Compliance review

The record is held with its clocks, its gaps and the systems they sit on.

Release · Append result · Send to compliance
Released — to the signer the state names
6Sample and inventory records updatedOnly where write access and records policy allow it
7Outcome evaluatedRecord completeness, clock coverage, compliance corrections and what review found
Corrections

Each compliance correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items
Certifying under 40 CFR 141.155(c).
Submitting anything to the primacy agency.
Deciding that a result is a violation.
Issuing public notification to customers.
Automation boundaryAgent acts unaided
Hold the sample, the method behind it and the holding.
Track each notification clock against its own trigger.
Name the rule version each monitoring schedule is running under today.
Reconcile the report against the monitoring data already submitted.
Nothing is certified or submitted except by a named person, inside agreed boundaries.
Judging whether a monitoring waiver applies.
Telling the state the system is in compliance.
Setting the material classification of a line.
Changes to sampling plans or inventory records.

Example output

One sample, annotated

Our wildfire, storm and water analytics agent returns a score and its error direction; this record is the compliance record behind one sample.

Compliance record · single sampleIllustrative example
Sample
Recorded as
System
Record on file
Confidence
Held for
Lead tap sample, first draw
Above the action level, unconfirmed
Community water system
Laboratory report, 11 August 2026
Held uncertified
The signer the primacy agency names
As receivedTaken from the laboratory report and the inventory entry — it reaches no further than those records.
What the record holds Method and holding time Averaging basis applied Inventory entry, dated
Why no violation callWhether a result is a violation is a primacy agency call, not a model output.
ActionReleaseAppend resultSend to compliance
What the score decidesA record under the configured threshold gets a compliance read before the signer sees it.

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 sampleFrom the system that drew it
03Record

Where the record is used

Our compliance reporting assistant works a register and names no regime; the mining discharge agent works wastewater out, and this page works drinking water in.

01Approved path

The state decides who signs

40 CFR 141.155(c) certifies not only that the report went out but that it is consistent with monitoring data already submitted.

02Human review

What was checked, and not found

Part 141 names no federal responsible official; who signs, whether a licence number appears and whether it is electronic are set by the state primacy agency.

04Build an evidence trail

The sample, the method behind it and the person the state named stay on the record.

Integrations

Typical integrations

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

Laboratory and samplingLIMS · certified laboratories
Chain-of-custody records
Inventory and asset recordsGIS · service line inventory
Material evidence records
Monitoring schedulesCompliance calendar · sampling plans
Rule version in force

Agent

Drinking water compliance record

Reads the samples
Assembles the record
Holds for the signer

Notification and reportingPrimacy agency portal · state reporting
Public notification records
Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six passes between the model and the signer

Six passes, every one narrower than the last. What is still there at the end is mapped below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model
L6Rollback / safe modeNarrow the agent to record assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and rule version; subpart I compliance falls on 1 November 2027, not before.Track
L4TraceabilityRecord each sample, the method behind it, the system it came from and every read.Record
L3Signer releaseHold the record for the person the state names; it governs release, not whether a limit was met.Gate
L2Policy guardrailsTest the record against Part 141 as your state configures it; the PFAS proposals stay proposed.Restrict
L1Confidence thresholdsRoute a thin record to a compliance read before the named signer sees it.Require review
Model coreRecord assembled — the sample, the system, the clocks and completeness
L1 – L2Test whether a record may stand
L3Puts the certification in a person's hands
L4 – L5Keep the sample and the method behind it
L6Holds the notice undrafted when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the record that comes out.

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

Surface — the record the state reads
Depth of coverage ▼
E1Final-output evaluationDid the record hold the method and holding time behind each result?
E2Step-level evaluationDid the agent read the right system, the right rule version and the live schedule?
E3Tool evaluationDid it read and write the correct sample record and the correct inventory entry?
E4Confidence calibrationDo low-confidence records actually attract more compliance corrections?
E5Slice evaluationHow does performance change across specific system types?
E6Business outcomeHow many records needed a correction before the signer released?
Floor — the outcome the system answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed where it first shows in the agent.

Agent lifecycleDirection of processing →
01 · Retrieval1 mode
JC-03

Stale schedule read

The schedule read is not the version now in force.

Stage gathersThe systems, the samples, the schedules and the clocks
02 · Reasoning2 modes
JC-04

Result judged on wrong basis

A result is read against the wrong averaging basis.

JC-06

Proposal read as final rule

A May 2026 PFAS proposal is worked as if in force.

Stage proposesThe samples, their versions and completeness
03 · Tool / write2 modes
JC-02

Thin record passed forward

A record moves on without the compliance read.

JC-05

Result bound to wrong system

A sample is filed against the wrong system.

Stage writesOnly where write access and approval policy allow it
04 · Output1 mode
JC-01

Certified, method unrecorded

The record shows a certification but not what supported it.

Stage returnsThe record a signer certifies and the state reads
05 · Change / Version1 mode
JC-07

Silent version regression

A rule update moves the schedule, not the record.

Stage tracksModel, prompt, version rules and record fields
Sev-1 · a record certified on no sample Sev-2 · a wrong result reaches the record Sev-3 · source degrades, record holds unsigned

Affected slices

Small systems absorb the corrections

A system-level version-currency figure can read clean while small community systems carry most of the corrections. Nestack reports the correction rate by system type, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example
SliceFailure rateLift Lift vs. thresholdStatus
Small community systems6.9%3.6× Review
Community systems above the size threshold4.9%2.6× Review
Non-transient non-community systems3.1%1.6× Watch
Systems with lead service lines1.3%0.7× Normal
Bar: correction-rate lift vs. lead-service-line system baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

What the wrong rule version costs

A cycle ends when the wrong rule version is a standing case. That suite is what the next record built is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect

Correction rate rises on small community systems.

02Diagnose

The sampling schedule running under a rule that another schedule had already replaced is read back until one cause remains.

03Improve

Every change leaves numbered, with the systems that caused it filed beneath.

04Verify

A single red system case is enough to keep the release back.

05Learn

The case is permanent, and the version rules are rewritten with it.

Learn → DetectThe return edge. The next record is measured against a suite 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, evidence assembly, evaluation, integration, then production validation and handover.

Workstream Week 1Week 2Week 3Week 4Week 5Week 6
01Drinking-water-clock discovery and boundary work.
02Laboratory, inventory and schedule sources.
03Rule-version-to-schedule and record-coverage mapping.
04Sample record ingestion.
05System, method and result binding.
06Completeness scoring and review routing.
07Signer certification workflow.
08Inventory and laboratory integration.
09Rule-version and sampling cases.
10Guardrails and primacy 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 system type, one year ProductionProduction compliance workflow AdvancedMultiple systems / states
Introduced at Pilot
Record assembly to your schedules
Named-signer release
Rule-version baseline
Introduced at Production
Reporting by system
Release workflow in your systems
Approved write-back
Laboratory-system integration
Introduced at Advanced
Multi-state primacy duties
Cross-rule record packs
Large sampling programmes
Multi-state primacy 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 systems and the schedules they run System mapping and sample captureWeek 1
02Representative laboratory and inventory records Record binding, clock logic and the sample baselineWeek 2
03Your primacy agency requirements and forms Schedule mapping, system binding and the automation boundaryWeek 1
04Access to relevant APIs, feeds or exports Laboratory, inventory and schedule-source assessment, then integration setupWeek 2
05Records you would not want submitted Version cases and the evaluation suiteWeek 4
06What no compliance record may conclude Completeness scoring, review routing, guardrails and release controlsWeek 3
07A named signer your state recognises Signer release 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

These bands measure real work and not drawing space, which is why the fifth week is doubled up.

Phase W1W2W3W4W5W6
Discovery W1
Build W2 – W3
Evaluate W4 – W5
Pilot & Launch W5 – W6
Week focus W1Compliance workflow discovery, schedule mapping and the automation boundary W2Source integration and the rule-version baseline W3Record assembly, clock logic and release controls W4Evaluation suite, version cases and failure-mode testing W5Record integration, pilot systems and targeted corrections W6One compliance year run under the operator the state names, then Agent Care handover
Reading the bandEach bar runs across only the weeks its own work is named for. The fifth week doubles because the work does.
At the end of W6When the compliance record validates, Agent Care assumes the watch.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Energy AI agent

Build a drinking water compliance agent around the record your state will ask for.

Show us one result and the schedule it belongs to, and we will tell you which rule version that schedule is on. Subpart I is effective 30 December 2024 and compliance falls 1 November 2027; oral argument is expected this autumn. Both PFAS rules of May 2026 remain proposals.

Nestack Agents · Drinking waterAGT-EN-15 · Agent Care available after launch