Main site

Travel & Hospitality AI agent · Aviation emissions units

Aviation Emissions Units AI Agent

Resolve CORSIA scope per state pair against the list ICAO republishes each year, hold every emissions unit with its programme, vintage and conditions, and run the four clocks the operator answers to.

Typical delivery
4–6 weeks
Deployment
Your stack
Named operator
Scope expires
After launch
Agent Care

What this agent does

Assembles the eligibility case, never the cancellation

In

  1. 01

    A reporting year closes, and the verified Emissions Report reaches the State by 30 April.

  2. 02

    A scaling factor publishes 31 October, and the State notifies the requirement by 30 November.

Reason

  1. 03

    A unit is offered, and the ICAO eligible-units document settles whether it qualifies at all.

  2. 04

    A unit carries a vintage, and only reductions dated 2021 to 2026 sit inside the window.

  3. 05

    A cancellation falls due 31 January 2028, and no first-phase deadline has yet run.

Decide

  1. 06

    A cancellation completes, and a verified Emissions Unit Cancellation Report follows by 30 April.

  2. 07

    A state pair is queried, and ICAO republishes the participating list before the next answer.

Out

  1. 08

    A Union airport is served, and Article 5 of Regulation (EU) 2023/2405 sets a yearly uplift share.

  2. 09

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

Product statement

The agent resolves scope and holds each unit with its conditions. An accredited verifier assesses it; the operator submits the report and cancels the units.

Example workflow

One state pair, unit to cancellation

  1. 1Unit and route records receivedRegistry statements, programme attestations, route schedules or fuel uplift records
  2. 2Eligibility context assembledThe unit, the programme that issued it, the vintage it carries and the state pair it answers for
  3. 3Submission state draftedThe units, their conditions and the scope answer
  4. 4Controls appliedEligibility checks, vintage checks, deadline checks and completeness confidence

No human action required

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

5DecisionSplits at the completeness gate
Eligibility settled

Goes to the operator to cancel.

Anything lapsed

Adds an environment read first.

Environment review

The submission is held with its units, the conditions on them and the scope behind it.

Cancel · Append record · Send to environment

Cancelled — by the operator

  1. 6Unit and route records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedEligibility accuracy, scope coverage, environment corrections and what review found

Corrections

Each environment correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Cancelling a unit in the registry.
    • Assessing a cancellation report for misstatement.
    • Submitting the verified Emissions Report to the State.
    • Deciding that a state pair falls outside scope.

  • Automation boundary

    Agent acts unaided

    • ✓Hold every unit with its programme and its vintage.
    • ✓Resolve every state pair against the list now in force at ICAO.
    • ✓Mark the unit whose eligibility has lapsed since it was bought.
    • ✓Track four submission clocks the operator has to meet.

    Nothing is cancelled or submitted except by the operator, inside the agreed boundaries.

    • Judging whether a unit still qualifies.
    • Telling a State that the fleet has offset.
    • Setting the scaling factor for a compliance year.
  • Changes to unit, route or fuel records.

Example output

One emissions unit, annotated

Our corporate travel policy agent reads a claim against policy; this record is what a single emissions unit carries.

Submission state · single unit

Illustrative example
Unit
Listed programme, one unit
Recorded as
Matched to the eligibility list in force
Vintage
First-phase window
Source of record
Registry statement, 3 August 2026
Confidence
Held uncancelled
Held for
The operator, as the instrument names it

As received

Taken from the registry statement and the programme attestation — it reaches as far as those sources do.

What the record holds

  • Registry statement
  • Programme attestation
  • Vintage window

Why no cancellation here

Whether a unit still qualifies is settled on the list, not by a model.

What the score decides

Below the configured threshold the submission picks up an environment read first.

Action

CancelAppend recordSend to environment

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 emissions unit

From the programme that issued it

03Submission

Where the record is used

Our vessel emissions agent splits one consumption dataset across four instruments for a designated company person; this page asks whether a unit still qualifies and whether the scope answer has changed since last year.

01Approved path

A unit, and its conditions

The ICAO eligible-units document reached its thirteenth edition, approved 31 October 2025, and every listed programme carries authorisation, corresponding-adjustment and vintage conditions.

02Human review

What was checked, and not found

Annex 16 Volume IV was never read directly — ICAO document pages refused the request, so every paragraph reaches us through ICAO model-regulations training material — and the offsetting weighting, the ReFuelEU synthetic sub-quota and the UK obligation step for 2026 remain unconfirmed.

04

Build an evidence trail

The unit, the programme that issued it and the person who cancelled 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.

  • Fuel and flight sources

    • Fuel uplift records
    • Flight and route schedules

  • Units and registries

    • Registry statements
    • Programme attestations

  • Verification and reporting

    • Verifier query log
    • Submitted report filings

Agent

Aviation emissions units

  • Reads the units
  • Resolves scope and vintage
  • Holds for the operator

  • Compliance schemes

    • CORSIA
    • EU ETS
    • UK ETS
    • ReFuelEU reports

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six conditions between the model and the operator

Six conditions, and a unit has to meet all of them. Whatever qualifies is charted in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeConfine the agent to eligibility assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and eligibility rules; the FAA CORSIA MRV Program is voluntary, and COM(2026) 616 final of 17 July 2026 is a proposal only.Track
L4TraceabilityRecord each unit, its programme, the state pair it answers for and every read of the submission.Record
L3Operator releaseHold the submission for the operator; that hold governs release, not whether a unit still qualifies.Gate
L2Vintage guardrailsTest every unit against the eligible-units document and its conditions; the first-phase vintage window shuts 31 December 2026.Restrict
L1Confidence thresholdsRoute a thin submission to an environment read before the operator sees it.Require review
Model coreSubmission assembled — the units, their conditions, the scope and completeness
L1 – L2
Test whether a submission may stand
L3
Leaves the cancellation with the operator
L4 – L5
Keep the unit and the programme behind it
L6
Holds the cancellation unmade when signals degrade

How Nestack evaluates it

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

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

Surface — the submission a verifier assesses

  1. E1Final-output evaluationDid the submission carry the conditions each unit has to meet?
  2. E2Step-level evaluationDid the agent read the right state pair, the right edition and the live record?
  3. E3Tool evaluationDid it read and write the correct unit and the correct vintage?
  4. E4Confidence calibrationDo low-confidence submissions actually attract more environment corrections?
  5. E5Slice evaluationHow does performance change across specific state pairs?
  6. E6Business outcomeHow many submissions needed a correction before the operator cancelled?

Floor — the obligation the operator owns

Failure modes

Where each failure originates in the agent

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

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • KA-03

      Stale eligibility read

      The list read is not the edition now in force.

    Stage gathers

    The units, the routes, the dates and the records

  2. 02 · Reasoning

    2 modes

    • KA-04

      Unit asserted, not shown

      A unit is called eligible without its conditions.

    • KA-06

      Proposal read as adopted

      An extension still in procedure is worked as law.

    Stage proposes

    The units, their conditions and the scope

  3. 03 · Tool / write

    2 modes

    • KA-02

      Thin submission passed on

      A submission moves without the environment read.

    • KA-05

      Unit bound to wrong state pair

      A record is filed against the wrong route.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • KA-01

      Cancelled, conditions unrecorded

      The report shows a cancellation but not what qualified it.

    Stage returns

    The submission a verifier assesses and a State reads

  5. 05 · Change / Version

    1 mode

    • KA-07

      Silent vintage regression

      A configuration change moves the window, not the record.

    Stage tracks

    Model, prompt, eligibility rules and unit fields

  • Sev-1 · a unit cancelled on no eligibility
  • Sev-2 · wrong scope answer reaches the State
  • Sev-3 · source degrades, submission holds
See the complete Travel & Hospitality failure-mode catalogue

Affected slices

State pairs newly in scope absorb the corrections

A state-pair-level scope-accuracy figure can read clean while the pairs that only just entered carry most of the rework. Nestack reports the correction rate by state pair, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
State pairs newly in scope8.9%3.6×Review
State pairs left off the list6.4%2.6×Review
Routes touching a Union airport4.0%1.6×Watch
Intra-EEA routes under the ETS1.5%0.6×Normal

Bar: correction-rate lift vs. intra-EEA route baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What a lapsed vintage costs

A cycle ends when the ineligible vintage is a standing case. That suite is what the next report assembled is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on state pairs newly in scope.

  2. 02

    Diagnose

    The unit bought in good faith that the eligibility list had already moved past is read back until one cause remains.

  3. 03

    Improve

    Changes leave numbered, and the submissions that forced them travel attached.

  4. 04

    Verify

    One submission case still failing is enough to stop the whole release.

  5. 05

    Learn

    The case is kept, and the eligibility rules are rewritten alongside it.

  6. Learn → Detect

    The return edge. The next submission is measured against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope set beside the delivery timeline. Week structure follows the six-week plan — discovery, sources, submission assembly, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Unit-eligibility and automation-boundary scoping.Week 1
  2. 02Fuel, route and registry sources.Week 1, Week 2
  3. 03State-pair scope and unit-condition coverage mapping.Week 1, Week 2
  4. 04Unit and route record ingestion.Week 2
  5. 05Unit, programme and vintage binding.Week 2, Week 3
  6. 06Eligibility scoring and review routing.Week 3
  7. 07Operator cancellation workflow.Week 3, Week 4
  8. 08Fuel and registry-system integration.Week 5
  9. 09Eligibility and scope cases.Week 4, Week 5
  10. 10Guardrails and cancellation controls.Week 4
  11. 11Submission-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 state pair, one yearProductionProduction offsetting workflowAdvancedMultiple schemes / entities
Introduced at Pilot
Submission assembly to your routes✓✓✓
Operator cancellation release✓✓✓
Unit-eligibility baseline✓✓✓
Introduced at Production
Reporting by state pair—✓✓
Cancellation workflow in your systems—✓✓
Approved write-back—✓✓
Fuel-record integration—✓✓
Introduced at Advanced
Multi-scheme operators——✓
Cross-scheme submission packs——✓
Large route networks——✓
Multi-scheme deadline controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, network size, 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 routes and the schemes each one owes

    Route mapping and unit-record capture

    Week 1

  2. 02Representative registry, fuel and verifier records

    Record binding, eligibility logic and the scope baseline

    Week 2

  3. 03Your monitoring plan as the State approved it

    Unit mapping, route binding and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Registry, fuel and verifier-source assessment, then integration setup

    Week 2

  5. 05Submissions you would not want verified

    Vintage cases and the evaluation suite

    Week 4

  6. 06What no emissions unit may guarantee

    Eligibility scoring, review routing, guardrails and release controls

    Week 3

  7. 07A nominated focal point for the operator

    Cancellation 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 bars here measure effort and not the page, and that is why two phases have to share one week.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Scheme discovery, route mapping and the automation boundary

W2Source integration and the unit-eligibility baseline

W3Submission assembly, eligibility logic and release controls

W4Evaluation suite, vintage cases and failure-mode testing

W5Registry integration, pilot submissions and targeted corrections

W6One compliance year run under the nominated focal point, then Agent Care handover

Reading the band
Each bar spans only the weeks its own work is named for. Two phases share the fifth because the work does.
At the end of W6
When the unit record validates, Agent Care picks the agent up.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Travel & Hospitality AI agent

Build an emissions units agent around the document that decides whether a unit still counts.

Show us one unit and the programme behind it. Not a measurement problem. A provenance one: the eligible-units document reached a thirteenth edition on 31 October 2025, and the first-phase cancellation deadline of 31 January 2028 has never run. Fuel buying is a different agent.

Typical delivery
4–6 weeks
Deployment
Your route network
Named operator
Scope expires
After launch
Agent Care available

Nestack Agents · Aviation emissions units

AGT-TH-18 · Agent Care available after launch

Build an emissions units agent

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

About you
Your route network

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