Main site

Energy AI agent · Sustainability

Sustainability-Reporting Copilot AI Agent

Assemble each disclosure figure with the source record and method behind it, apply the framework mapping you have configured, and hold the workpaper for the named preparer, who decides what is published.

Typical delivery
4–6 weeks
Deployment
Your stack
Preparer signs
Source-linked
After launch
Agent Care

What this agent does

Assembles the evidence, not the sign-off

In

  1. 01

    Ingesting meter data, fuel and invoice records and factor sets from supported energy, ERP or data-platform sources.

  2. 02

    Normalising units, periods and site names, and carrying each datapoint forward with the record it came from.

Reason

  1. 03

    Assembling the workpaper behind each figure — source record, method, factor version and period.

  2. 04

    Applying the framework mapping, boundary and consolidation approach the organisation has configured.

  3. 05

    Reconciling each total against the prior period and writing the movement against the datapoints that moved it.

Decide

  1. 06

    Marking a market-based figure with the instruments it rests on and whether they have been retired.

  2. 07

    Routing the assembled figure and its narrative to the named preparer for acceptance.

Out

  1. 08

    Retaining the datapoint, the method, the preparer's amendments and the acceptance in the workpaper.

  2. 09

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

Product statement

The agent assembles the basis; the named preparer decides what is signed, and the organisation stays the reporting entity.

Example workflow

One figure, meter read to sign-off

  1. 1Datapoint receivedMeter record, fuel or invoice line, factor set or an ERP extract
  2. 2Basis assembledSource record, period, boundary, method and factor version, each named
  3. 3Figure proposedFigure, movement, dual basis and confidence
  4. 4Controls appliedTraceability checks, the configured mapping rules, instrument checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is filed or published at any of them — the agent is assembling evidence, and the preparer's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the named preparer to accept.

Low confidence

Adds a data-owner check first.

Preparer sign-off

The figure is held with its source record, its method and the confidence.

Accept · Amend · Send to data owner

Accepted — released to report

  1. 6Reporting systems updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedAmendment rate, restatements, evidence gaps and findings raised on review

Amendments

Every preparer amendment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Publishing or filing a disclosure of any kind.
    • Signing off a figure or a restated prior total.
    • Making an environmental marketing claim of any kind.
    • Deciding that a framework applies to the entity.

  • Automation boundary

    Agent acts unaided

    • ✓Collect each datapoint with its source record and method.
    • ✓Apply the framework mapping the organisation has configured.
    • ✓Reconcile against the prior period and explain the movement.
    • ✓Hold the assembled workpaper for the named preparer.

    Any write happens inside the boundaries agreed at implementation, never ahead of sign-off.

    • Changing a method, a boundary or an emissions factor.
    • Retiring or allocating a market-based instrument.
    • Stating a carbon-neutral or offset-based position.
  • Changes to mapping, thresholds or approval rules.

Example output

One figure in the workpaper, annotated

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

Disclosure output · single datapoint

Illustrative example
Site
Northern network depot
Datapoint
Purchased electricity for the reporting period
Value
412,600 kWh
Source of record
Meter register on file
Confidence
93%
Method
Location and market basis

As received

Taken from the meter register and the supplier invoice — nothing on this side is written by the agent.

Evidence used

  • Meter register read
  • Supplier invoice
  • Factor set version

Why this method

It is the method configured for this boundary — not chosen per datapoint.

What the score decides

Below the configured threshold the figure picks up a data-owner check before it.

Action

AcceptAmendSend to data owner

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 datapoint

From the source record

03Assembly

Build from the record

Draw on the meter, invoice and factor records and the mapping the organisation has configured.

01Approved path

Every figure keeps its source

Routine datapoints arrive already assembled with their basis attached.

02Human review

Send review to the thin evidence

Evidence gaps, unretired instruments and low-confidence figures are marked, so the preparer's read starts where the basis is weakest.

04

Build an evidence trail

The datapoint, the method applied to it and the preparer who accepted it stay in the workpaper.

Integrations

Typical integrations

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

  • Energy and emissions data

    • Meter data
    • SCADA historians
    • Fuel, fleet and invoice records

  • Sustainability platforms

    • Watershed
    • Persefoni
    • Sphera
    • Net Zero Cloud

  • Enterprise systems

    • SAP
    • Oracle
    • Workday
    • Snowflake
    • Databricks

Agent

Sustainability reporting

  • Reads the records
  • Assembles the basis
  • Holds for sign-off

  • Reporting and workflow

    • Workiva
    • disclosure tooling
    • Evidence and workpaper stores

  • 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 disclosure

Each control encloses the last. Whatever slips past all of them is named below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modePull the agent back to data collection when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, mapping and factor-set version changes.Track
L4TraceabilityRecord the source, the method, the factor version, the amendments and the acceptance.Record
L3Preparer sign-offHold figures for the named preparer; it governs release, not whether an accepted figure is right.Gate
L2Policy guardrailsTest figures against the configured mapping and claim rules; a failure returns the figure.Restrict
L1Confidence thresholdsRoute low-confidence figures to a data-owner check before the preparer sees them.Require review
Model coreFigure proposed — value, source record, method, movement and confidence
L1 – L2
Test whether a figure may stand
L3
Puts the release in a preparer's hands
L4 – L5
Keep the datapoint and the method applied
L6
Cuts back to data collection when signals degrade

How Nestack evaluates it

Evaluate the assembly workflow — not only the finished disclosure.

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

Surface — the disclosure the reader sees

  1. E1Final-output evaluationDid each figure trace back to a source record and a stated method?
  2. E2Step-level evaluationDid the agent use the right records, mapping and factor versions?
  3. E3Tool evaluationDid it read the correct system, site and reporting period?
  4. E4Confidence calibrationDo low-confidence figures actually attract more preparer amendments?
  5. E5Slice evaluationHow does performance change across specific disclosure areas?
  6. E6Business outcomeHow many figures needed an amendment or a restatement afterwards?

Floor — the outcome the organisation answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, placed at the stage each one originates.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • LQ-03

      Superseded factor set

      An emissions factor read from a retired version.

    Stage gathers

    Meter records, invoices , with the source each came from

  2. 02 · Reasoning

    2 modes

    • LQ-04

      Untraceable figure

      A value no source record or repeatable method produces.

    • LQ-06

      Silent method drift

      Boundary or factor vintage changes without a note.

    Stage proposes

    The figure, the method behind it and confidence

  3. 03 · Tool / write

    2 modes

    • LQ-02

      Unsupported explanation

      A movement is given a cause the evidence cannot carry.

    • LQ-05

      Unretired instrument

      A market-based figure rests on an unretired instrument.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • LQ-01

      Marketing claim in the draft

      A neutrality or renewable claim written unbidden.

    Stage returns

    The disclosure the preparer signs and the reader relies on

  5. 05 · Change / Version

    1 mode

    • LQ-07

      Silent factor regression

      A model or factor-set change moves a figure without a note.

    Stage tracks

    Model, prompt, mapping and factor-set changes

  • Sev-1 · a figure that cannot be traced
  • Sev-2 · a wrong figure reaches the draft
  • Sev-3 · source degrades, figure routes to review
See the complete Energy & Utilities failure-mode catalogue

Affected slices

Overall quality can hide one weak area

Scope 3 value-chain estimates absorb most of the amendments, and a clean overall figure will not show you that. Nestack reports the preparer-amendment rate by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Scope 3 value-chain estimates6.1%3.7×Review
Market-based instrument claims4.3%2.6×Review
Newly acquired or divested sites3.3%2.0×Watch
Metered Scope 1 and 2 data1.3%0.8×Normal

Bar: preparer-amendment-rate lift vs. metered Scope 1 and 2 baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on a test, not a note

The cycle does not end in a post-mortem. It ends in a case the next release must clear, and that suite is what the next disclosure is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Preparer-amendment rate rises in a disclosure area.

  2. 02

    Diagnose

    Which datapoints moved, and under which method? The workpapers are read back until the cause narrows to one.

  3. 03

    Improve

    Each change carries a version and the disclosures that forced it.

  4. 04

    Verify

    Release waits on the affected cases going green again.

  5. 05

    Learn

    The case is kept for good, and the method notes are rewritten.

  6. Learn → Detect

    The return edge. The next detection runs 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, assembly workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Disclosure workflow discovery and automation-boundary definition.Week 1
  2. 02Meter, ERP and data-platform assessment.Week 1, Week 2
  3. 03Framework mapping, boundary and consolidation approach.Week 1, Week 2
  4. 04Datapoint ingestion and normalisation.Week 2
  5. 05Workpaper assembly and source binding.Week 2, Week 3
  6. 06Confidence scoring and gap routing.Week 3
  7. 07Preparer sign-off workflow.Week 3, Week 4
  8. 08Data-platform and reporting integration.Week 5
  9. 09Audit-evidence regression cases.Week 4, Week 5
  10. 10Guardrails and claim controls.Week 4
  11. 11Workpaper-trail 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 reporting entity, one periodProductionProduction data platformsAdvancedMultiple entities / mappings
Introduced at Pilot
Assembly to your records and mapping✓✓✓
Preparer sign-off✓✓✓
Evidence-quality baseline✓✓✓
Introduced at Production
Reporting by disclosure area—✓✓
Preparer review in your systems—✓✓
Approved write-back—✓✓
Data-platform integration—✓✓
Introduced at Advanced
Multi-framework mapping rules——✓
Multi-stage disclosure approvals——✓
High datapoint volume——✓
Multi-framework disclosure controls——✓
Build priceFrom $5,000From $8,000Custom 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.

  1. 01Your reporting boundary and entity structure

    Datapoint ingestion and source-record mapping

    Week 1

  2. 02Representative prior-period disclosures

    Assembly baseline, method extraction and source binding

    Week 2

  3. 03Your framework mapping and its named owner

    Framework mapping, boundary and consolidation approach

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Meter, ERP and data-platform assessment, then integration setup

    Week 2

  5. 05Claims you would not want published

    Evidence cases and failure-mode testing

    Week 4

  6. 06What a draft may never assert on its own

    Confidence scoring, gap routing, guardrails and claim controls

    Week 3

  7. 07Named preparers to accept figures

    Preparer sign-off 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

The bands sit on the weeks the work occupies, which is why the fifth carries two kinds at once.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Disclosure workflow discovery, mapping and the automation boundary

W2Source integration and the assembly baseline

W3Assembly workflow, confidence logic and sign-off controls

W4Evaluation suite, claim checks and failure-mode testing

W5Reporting-system integration, pilot areas and targeted corrections

W6One disclosure cycle drafted under the reporting team, then handover

Reading the band
A band covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.
At the end of W6
The cycle closes the validation and Agent Care takes over monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Energy AI agent

Build a sustainability agent around your workpaper, not around your report template.

Show us your datapoints, the records behind them and the mapping you report against. The preparer named in your reporting policy is the one who signs, and we will name what never leaves their desk.

Typical delivery
4–6 weeks
Deployment
Your data and reporting stack
Preparer signs
Source-linked
After launch
Agent Care available

Nestack Agents · Sustainability reporting

AGT-EN-05 · Agent Care available after launch

Build a sustainability agent

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

About you
Your reporting

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