Travel & Hospitality AI agent · Aviation testing
Aviation Testing Programme AI Agent
Tally every safety-sensitive name into the pool with the date it entered or left, tie each test result to the pool it came from, and hold the report for the Designated Employer Representative.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Named DER
- March deadline
- After launch
- Agent Care
What this agent does
Assembles the report, never the certification
In
- 01
A test is ordered, and § 120.105 sets which duties put an employee in the pool.
- 02
A reporting year closes, and § 120.119 puts the MIS report with the FAA by 15 March.
Reason
- 03
A report is prepared, and § 40.26 names the form at appendix J to 49 CFR part 40.
- 04
A year of results is certified, and § 120.109(b) and § 120.217(c) set next year off it.
- 05
A rate notice publishes, and 90 FR 56249 of 5 December 2025 governs the current year.
Decide
- 06
A collection is ordered on oral fluid, and two HHS-certified laboratories are required first.
- 07
An oral fluid test cannot run, and 91 FR 25507 of 11 May 2026 finds no laboratory certified.
Out
- 08
A return-to-duty case opens, and 91 FR 10518 of 4 March 2026 is a notification, not a rule.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent holds the pool census and reconciles the tests to it. The Designated Employer Representative certifies the accuracy and completeness of the MIS report.
Example workflow
One pool month, evidence to certification
- 1Pool and test evidence receivedRoster feeds, selection lists, collection records or laboratory results
- 2Pool context assembledThe employee, the duty performed, the date the name entered the pool and the date it left
- 3Report evidence draftedThe tests reconciled, the pool months behind them, the gaps and completeness
- 4Controls appliedCensus checks, reconciliation checks, reporting-clock 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.
Goes to the Designated Employer Representative to certify.
Adds a compliance analyst read first.
Compliance review
The report is held with its pool months, its unreconciled tests and the categories they fall in.
Certify · Append evidence · Send to compliance
Certified — by the named representative
- 6Pool and test records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedCensus accuracy, test reconciliation, analyst corrections and what the read 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
- Certifying the MIS report under § 120.119.
- Deciding that a duty is not safety-sensitive.
- Submitting the annual report to the FAA.
- Removing a name from the testing pool.
Automation boundary
Agent acts unaided
- ✓Hold the pool census with the date every name entered.
- ✓Reconcile every test result to the pool month that produced it.
- ✓Track the reporting clock against the 15 March date that closes it.
- ✓Flag the month whose pool census is not evidenced.
Nothing is certified or submitted except by a named person, inside the agreed boundaries.
- Judging whether the report may be certified.
- Telling the FAA a testing programme is in order.
- Setting the rate a testing programme must meet.
- Changes to pool, roster or test records.
Example output
One pool month, annotated
The reporting deadline falls on 15 March; this record is what a single pool month carried into the report.
Report evidence · single pool month
Illustrative example- Pool month
- Random selection, § 120.109
- Recorded as
- Reconciled to the pool, census evidenced
- Test category
- Random drug
- Evidence of record
- Collection record, 3 August 2026
- Confidence
- Held uncertified
- Held for
- The Designated Employer Representative
As received
Taken from the roster feed and the collection record — it reaches as far as those sources do.
What the record holds
- Pool census date
- Collection record
- Laboratory result
Why no certification here
Whether the report may be certified is a § 120.119 act, not a model output.
What the score decides
Below the configured threshold the report picks up a compliance read before the DER sees it.
Action
CertifyAppend evidenceSend 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 test
From the pool that produced it
Where the evidence is used
Our driver qualification and Clearinghouse agent runs per-driver queries on each driver's own twelve-month clock; this page is a per-company annual certified aggregate that sets the following year's random rate, a feedback loop with no FMCSA analogue.
This year sets next year
What the representative certifies in March is what § 120.109(b) and § 120.217(c) work from: 90 FR 56249 of 5 December 2025 was signed by Brett A. Wyrick, Deputy Federal Air Surgeon.
What was checked, and not found
Retention under § 120.219 was not retrieved; the certification block on appendix J itself was not read; the status of the October 2024 electronic-signatures NPRM is unknown; and a laboratory certified after 11 May 2026 could not be ruled out.
Build an evidence trail
The test, the pool it was drawn from and the representative who certified stay on file.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Roster and duty records
- HR system
- duty rosters
- Safety-sensitive duty codes
Collection and laboratory
- Collection sites
- custody forms
- Laboratory results
Random selection
- Consortium
- C/TPA feeds
- Selection lists and draws
Agent
Aviation testing programme
- Reads the pool
- Reconciles the tests
- Holds for the DER
Reporting and returns
- MIS submissions
- exports
- Operations specification
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six audits between the model and the representative
Six audits of the same pool, and no two are alike. Whatever counts is set out in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a report may stand
- L3
- Puts the certification in a person's hands
- L4 – L5
- Keep the test and the pool behind it
- L6
- Stops at pool assembly when signals degrade
How Nestack evaluates it
Evaluate the whole assembly — not only the report evidence that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the report the FAA reads
- E1Final-output evaluationDid the evidence record what each pool month actually held?
- E2Step-level evaluationDid the agent read the right roster, the right month and the live selection list?
- E3Tool evaluationDid it read and write the correct pool and the correct test result?
- E4Confidence calibrationDo low-confidence reports actually attract more compliance corrections?
- E5Slice evaluationHow does performance change across specific test categories?
- E6Business outcomeHow many reports needed a correction before the representative certified?
Floor — the report the employer answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, each set where it first shows in the run.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- KC-03
Stale roster read
The roster read is not the pool now in force.
Stage gathers
The people, the tests, the pools and the dates
02 · Reasoning
2 modes
- KC-04
Report asserted, not shown
A month is called evidenced without its census.
- KC-06
Proposed panel read as live
The fentanyl NPRM is worked as though final.
Stage proposes
The tests, their pools and completeness
03 · Tool / write
2 modes
- KC-02
Thin report passed forward
A report moves on without the compliance read.
- KC-05
Test bound to wrong pool
A result is filed against the wrong month.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- KC-01
Certified, evidence unrecorded
The report shows a certification but not its basis.
Stage returns
The report a representative certifies and the FAA reads
05 · Change / Version
1 mode
- KC-07
Silent census regression
A configuration change moves the pool, not the report.
Stage tracks
Model, prompt, pool rules and report fields
- Sev-1 · a report certified on no census
- Sev-2 · wrong test reaches the report
- Sev-3 · source degrades, report uncertified
Affected slices
Random selections absorb the corrections
A pool-level census-accuracy figure can read clean while random selections carry most of the rework. Nestack reports the correction rate by test category, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Random drug selections | 7.7% | 3.6× | Review | |
| Pre-employment tests | 5.5% | 2.6× | Review | |
| Post-accident tests | 3.5% | 1.6× | Watch | |
| Follow-up and return-to-duty | 1.4% | 0.7× | Normal |
Bar: correction-rate lift vs. follow-up baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a mis-sized pool costs
A cycle closes when the mis-sized pool is a regression 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
- 01
Detect
Correction rate rises on random selections.
- 02
Diagnose
The headcount submitted in March that set a testing rate for the whole of the next year is read back until one cause remains.
- 03
Improve
The change ships numbered, and the tests that forced it ride with it.
- 04
Verify
Nothing releases while one touched pool case is still red.
- 05
Learn
It is retained for good, and the pool rules are amended in that same commit.
Learn → Detect
The return edge. The next testing year 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, report assembly, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Safety-sensitive pool discovery and boundary work.Week 1
- 02Roster, collection and result sources.Week 1, Week 2
- 03Test-to-pool and reporting-clock coverage mapping.Week 1, Week 2
- 04Test and pool-census ingestion.Week 2
- 05Employee, duty and pool binding.Week 2, Week 3
- 06Completeness scoring and review routing.Week 3
- 07Representative certification workflow.Week 3, Week 4
- 08Roster and testing-system integration.Week 5
- 09Pool-size and rate cases.Week 4, Week 5
- 10Guardrails and representative controls.Week 4
- 11Test-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 pool, one testing year | ProductionProduction reporting workflow | AdvancedMultiple pools / employers |
|---|---|---|---|
| Introduced at Pilot | |||
| Report assembly to your pools | ✓ | ✓ | ✓ |
| Representative release | ✓ | ✓ | ✓ |
| Safety-sensitive pool baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by test category | — | ✓ | ✓ |
| Certification workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Collection-site integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-employer consortia | — | — | ✓ |
| Cross-year evidence packs | — | — | ✓ |
| Large safety-sensitive rosters | — | — | ✓ |
| Multi-employer pool controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, pool 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.
You bring
We build with it
01Your safety-sensitive employees and the duties they hold
→Pool inventory mapping and census capture
Week 102Representative roster, collection and result records
→Test binding, pool logic and the census baseline
Week 203Your duty list under § 120.105 and § 120.215
→Pool mapping, test binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Roster, collection and result-source assessment, then integration setup
Week 205Pools you would not want counted
→Rate cases and the evaluation run
Week 406What no testing report may establish
→Completeness scoring, review routing, guardrails and release controls
Week 307A Designated Employer Representative to certify
→Certification 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 span is a week the work really needs and not a drawing, so one week ends up carrying a pair.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Testing programme discovery, pool mapping and the automation boundary
W2Source integration and the census-accuracy baseline
W3Report assembly, reconciliation logic and release controls
W4Evaluation suite, category cases and failure-mode testing
W5Record integration, pilot pools and targeted corrections
W6One testing year run under the designated representative, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its own work is named for. The fifth week carries two because the work does.
- At the end of W6
- Validation closes on live pools, and Agent Care picks up the watch.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Travel & Hospitality AI agent
Build a testing programme agent around the report your representative has to certify.
Show us one pool month and the tests drawn from it. Send us the census behind your last March filing and we will show you where the evidence for it stops. Oral fluid is authorised in the regulation and unusable in practice; hazard registers and safety policy are another agent.
- Typical delivery
- 4–6 weeks
- Deployment
- Your testing pool
- Named DER
- March deadline
- After launch
- Agent Care available
Nestack Agents · Aviation testing
AGT-TH-20 · Agent Care available after launch