Energy AI agent · Pipeline records
Pipeline Integrity Records AI Agent
Bind each entity clock in 49 CFR Parts 191, 192 and 195 to the record it consumed — the confirmed discovery, the thirty-day report, the fifteen-month manual review. No officer certifies any of it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- No signer
- Clock-bound
- After launch
- Agent Care
What this agent does
Holds the record, never files it
In
- 01
A discovery is confirmed, and 49 CFR 191.5(a) leaves no more than one hour to telephone the notice.
- 02
A liquid accident is called in, and § 195.52 wants revised estimates or confirmation within 48 hours.
Reason
- 03
A detection is logged, and § 191.15 allows 30 days for the report; § 195.54(b) adds a supplemental.
- 04
A manual review falls due, and § 192.605(a) caps the interval at 15 months, once each calendar year.
- 05
A gas annual report closes, and § 191.11 still says 15 March — the move to June was withdrawn.
Decide
- 06
An evaluation falls due, and § 192.805 leaves the interval to the operator plan, not to Part 192.
- 07
A reassessment is scheduled, and § 192.939 says 7 calendar years; § 195.452(j)(3) says 68 months.
Out
- 08
A distribution programme ages, and § 192.1007 wants re-evaluation at least every five years.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent assembles and tracks entity clocks. No officer certifies under Parts 191 to 195 — the operations manager releases, and the operator answers.
Example workflow
One clock, discovery to release
- 1Clock record receivedDiscovery times, inspection results, qualification records or report drafts
- 2Clock context assembledThe segment, the section setting the deadline and whether repair is immediate, 60-day or 180-day
- 3Record set draftedThe clock, its records, the gaps and completeness
- 4Controls appliedDeadline checks, record-sufficiency checks, date checks and completeness confidence
No human action required
Stages 1 to 4 run unaided, and nothing is filed at any of them — the agent is assembling, and the compliance lane opens at the completeness gate.
Goes to the operations manager to release.
Adds a pipeline compliance read first.
Compliance review
The set is held with its clock, its gaps and the segments they sit on.
Release · Append record · Send to compliance
Released — by the operations manager
- 6Segment and qualification 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 required
8 items
- Determining that an event is reportable.
- Telephoning the National Response Center.
- Filing an incident, accident or annual report.
- Setting the intervals in the qualification plan.
Automation boundary
Agent acts unaided
- ✓Hold the record each entity clock consumes, with the date it carries.
- ✓Track each clock against the section of Part 191, 192 or 195 that sets it.
- ✓Flag the 48-hour liquid follow-up that compliance calendars routinely omit.
- ✓Read the re-evaluation interval out of the operator qualification plan.
Nothing is reported or filed except by a named person, inside the agreed boundaries.
- Judging whether a repair condition was met.
- Telling PHMSA the operator is in compliance.
- Choosing the reassessment method for a segment.
- Changes to pressure, alignment sheets or field records.
Example output
One clock, annotated
Our grid asset maintenance copilot holds a finding for the named engineer; this record is what one entity clock consumed.
Record set · single clock
Illustrative example- Clock
- Written report, § 191.15
- Recorded as
- Drafted from the detection date, not filed
- Segment
- Regulated onshore gathering
- Record on file
- Operator detection log, 6 August 2026
- Confidence
- Held unfiled
- Held for
- The operations manager, by name
As received
Taken from the detection log and the incident file — it reaches as far as those sources do.
What the record holds
- Confirmed discovery time
- Who confirmed it
- Supplemental, dated
Why no reportability call
Reportability is a § 191.15 judgement for the operator, not a model output.
What the score decides
Under the configured threshold a compliance read comes first, and only then the manager.
Action
ReleaseAppend recordSend to complianceValue
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 clock
From the segment that owes it
Where the record is used
Our compliance reporting assistant works a register and names no regime, and our control room copilot works operations; this page is one regime and its statutory records.
No officer signs any of it
§ 192.939 states one maximum reassessment interval of 7 calendar years; the stress-tiered figures come from ASME B31.8S, incorporated by reference.
What was checked, and not there
No named-officer certification appears anywhere in the pipeline safety regulations, and no leak detection final rule was confirmed as published.
Build an evidence trail
The record, the clock it started and the person who confirmed discovery stay on the file.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Reporting and notification
- NRC hotline
- PHMSA portal
- Incident and accident reports
Integrity management
- ILI
- direct assessment
- Reassessment records
Operator qualification
- Qualification plan
- training
- Evaluation records
Agent
Pipeline integrity records
- Reads the clocks
- Assembles the record
- Holds for the manager
Work and asset systems
- GIS
- work management
- Leak, repair and survey records
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 file
Six layers that tighten going inward. Whatever reaches the centre is described in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a set may stand
- L3
- Puts the release in a person's hands
- L4 – L5
- Keep the record and the clock behind it
- L6
- Holds the report undrafted when signals degrade
How Nestack evaluates it
Evaluate the whole assembly — not only the record set that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the set an inspector reads
- E1Final-output evaluationDid the set record what each clock actually consumed?
- E2Step-level evaluationDid the agent read the right segment, the right section and the live plan?
- E3Tool evaluationDid it read and write the correct segment record and the correct clock?
- E4Confidence calibrationDo the sets marked low-confidence in fact draw more compliance corrections?
- E5Slice evaluationHow does performance change across specific segment types?
- E6Business outcomeHow many sets needed a correction before the manager released?
Floor — the outcome the operator answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, each placed at the stage in the agent where it starts.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- JB-03
Stale interval read
The plan read is not the revision now in force.
Stage gathers
The segments, the clocks, the records and the dates
02 · Reasoning
2 modes
- JB-04
Deadline asserted, not shown
A clock is called met without its record.
- JB-06
Superseded rule read as live
§ 192.478 was removed; the June date was withdrawn.
Stage proposes
The clocks, their records and completeness
03 · Tool / write
2 modes
- JB-02
Thin set passed forward
A set moves on without the compliance read.
- JB-05
Record bound to wrong segment
A record is filed against the wrong segment.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- JB-01
Released, record unkept
The file shows release but not what supported it.
Stage returns
The set a manager releases and an inspector reads
05 · Change / Version
1 mode
- JB-07
Silent interval regression
A plan revision moves the interval, not the record.
Stage tracks
Model, prompt, clock rules and record fields
- Sev-1 · a clock released on no record
- Sev-2 · a wrong segment reaches the file
- Sev-3 · source degrades, set holds unreleased
Affected slices
Gathering lines carry the corrections
A segment-level interval-provenance figure can read clean while regulated onshore gathering carries most of the corrections. Nestack reports the correction rate by segment, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Regulated onshore gathering | 7.4% | 3.6× | Review | |
| Gas distribution mains and services | 5.3% | 2.6× | Review | |
| Hazardous liquid line pipe | 3.3% | 1.6× | Watch | |
| Gas transmission in HCAs | 1.7% | 0.8× | Normal |
Bar: correction-rate lift vs. HCA gas transmission baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a borrowed interval costs
A cycle closes when the borrowed interval is a regression case. That suite is what the next record kept is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on regulated onshore gathering.
- 02
Diagnose
The re-evaluation cycle everybody quoted and nobody could find in § 192.805 is read back until one cause remains.
- 03
Improve
Changes go out numbered, and the segments that drove them ride along.
- 04
Verify
One segment case still failing keeps the release where it is.
- 05
Learn
It joins the suite for good, and the interval rules move in the same commit.
Learn → Detect
The return edge. The next clock 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
- 01Entity-clock discovery and automation-boundary work.Week 1
- 02Detection, inspection and plan sources.Week 1, Week 2
- 03Clock-to-section and segment-record-coverage mapping.Week 1, Week 2
- 04Clock record ingestion.Week 2
- 05Segment, section and record binding.Week 2, Week 3
- 06Completeness scoring and review routing.Week 3
- 07Manager release workflow.Week 3, Week 4
- 08Work and asset-system integration.Week 5
- 09Interval-source and clock cases.Week 4, Week 5
- 10Guardrails and record-keeping controls.Week 4
- 11Record-trail instrumentation.Week 5, Week 6
- 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 tier | PilotOne segment type, one year | ProductionProduction compliance workflow | AdvancedMultiple segment types / operators |
|---|---|---|---|
| Introduced at Pilot | |||
| Record assembly to your clocks | ✓ | ✓ | ✓ |
| Operations manager release | ✓ | ✓ | ✓ |
| Interval-provenance baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by segment | — | ✓ | ✓ |
| Release workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Work-management integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-operator registrations | — | — | ✓ |
| Cross-part record packs | — | — | ✓ |
| Large segment inventories | — | — | ✓ |
| Multi-mode deadline controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom 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.
You bring
We build with it
01Your segments and the parts they fall under
→Segment mapping and clock capture
Week 102Representative detection, inspection and qualification records
→Record binding, clock logic and the interval baseline
Week 203Your qualification plan and the individuals § 192.807 records
→Clock mapping, segment binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Detection, inspection and plan-source assessment, then integration setup
Week 205Records you would not want relied on
→Interval cases and failure testing
Week 406What no operator record may settle
→Guardrails, release controls, completeness scoring and review routing
Week 307A named operations manager to release
→Manager release 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 band is the honest length of one phase in working weeks, and that is why a week carries two.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Clock discovery, segment mapping and the automation boundary
W2Source integration and the interval-provenance baseline
W3Record assembly, clock logic and release controls
W4Interval cases, failure-mode testing and the evaluation suite
W5Record integration, pilot segments and targeted corrections
W6One reporting year run under the operations manager, then Agent Care handover
- Reading the band
- Each bar runs only across the weeks its own work is named for. The fifth week doubles because the work does.
- At the end of W6
- Once the operator record validates, Agent Care picks up the agent.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Energy AI agent
Build a pipeline records agent around the clocks Parts 191 to 195 already run.
Show us one confirmed discovery and what it started. Who certifies the annual report? No one — no officer is named, so the record answers alone. The dam safety agent tracks a person and a signature; this one tracks entity clocks, and deregulation has not relaxed enforcement.
- Typical delivery
- 4–6 weeks
- Deployment
- Your segment inventory
- No signer
- Clock-bound
- After launch
- Agent Care available
Nestack Agents · Pipeline records
AGT-EN-14 · Agent Care available after launch