Keep aviation knowledge grounded and inference selective
Airlines and aviation operators could retrieve approved procedures before paying a model to interpret them. This thesis keeps operational authority with qualified staff and reserves premium reasoning for permitted, non-safety-critical analysis.
Industry thesis — modelled, not deployed. Figures are potential savings.
5.3%Passenger demand growth in 2025IATA, January 2026
83.6%Record passenger load factor, 2025IATA, January 2026
4.3bnScheduled passengers in 2023ICAO, 2024
$23.0bnForecast 2026 airline net profitIATA, June 2026
Airlines & aviation — the cheapest trustworthy source answers first; frontier reasoning is paid for only when it earns its price.
Where the money goes today
Airlines & aviation: the cost of asking
IATA's full-year traffic release in January 2026 reported that passenger demand, measured in revenue passenger kilometres, rose 5.3% in 2025. It also reported a record annual passenger load factor of 83.6%, leaving little room for avoidable operational friction.
ICAO's 2024 reporting on 2023 air transport results recorded approximately 4.3 billion passengers on scheduled services. That passenger volume creates a broad service workload, but it is not a proxy for the number of questions suitable for AI.
IATA's June 2026 outlook cut the industry net-profit forecast for 2026 to $23.0 billion, roughly halving expected profitability on fuel prices and Middle East disruption. Those were forecast figures, and they frame why small operating costs merit scrutiny without establishing any Qua savings.
Three pressures
What makes this industry different
Disruption increases repetitive demand
Weather and network disruption create bursts of policy and service questions. Many need an approved answer rather than fresh reasoning.
Manual revisions change answers
Procedure applicability depends on operator, fleet and revision. Retrieval must preserve that context rather than blend incompatible instructions.
Safety authority stays human
Generated text must not become a maintenance release or flight instruction. Analysis and administrative support need clear boundaries from safety decisions.
Five applications
How Qua would run inside airlines & aviation
Each application is a real workflow, mapped to the tier that could answer it. The waterfall tries the cheapest trustworthy source first and only pays a frontier model when the expected value clears the gate.
Baggage allowance policy answers
Resolves at Personal Knowledge~20,000 questions/month
Service agents could reuse approved explanations for common allowance questions. Fare, route and carrier exceptions would require a more specific source check.
Potential saving — 90–100% of query spend.
Apply carrier, channel and customer-data policies before routing.
Match the itinerary context to a verified, current baggage-policy answer.
Return the cached answer and receipt; escalate interline or policy exceptions.
Ground handling procedure lookup
Resolves at Enterprise Search~25,000 questions/month
Teams could locate the right document section faster. The approved procedure would remain authoritative, with operational checks unchanged.
Potential saving — 80–100% of query spend.
Enforce station, operator and document-access permissions.
Check verified answers, then search current approved ground-handling documents.
Return the applicable source passage and receipt without external inference.
Maintenance record evidence retrieval
Resolves at Enterprise Search~15,000 questions/month
Records teams could assemble evidence for audit and planning without exporting technical files. The workflow would not diagnose faults or issue a release to service.
Potential saving — 80–100% of query spend.
Apply aircraft-record entitlements and clamp the workflow to perimeter retrieval.
Check verified references, then retrieve authorised maintenance records.
Return document identifiers, source passages and a receipt for qualified review.
Passenger disruption message drafting
Resolves at Fast Models~30,000 questions/month
Service teams could adapt approved facts for different communication channels. Staff would check entitlement wording and approve publication.
Potential saving — 80–95% of query spend.
Mask passenger identifiers and apply approved-provider restrictions.
Retrieve confirmed disruption facts and approved assistance-policy wording.
Use a permitted Fast Model to draft a message and attach its receipt.
Post-disruption commercial scenario review
Resolves at Pro Models~10,000 questions/month
Commercial planners could compare historical recovery options and their service trade-offs. The output would not dispatch aircraft, assign crews or replace certified operational systems.
Potential saving — 0–20% of query spend.
Restrict inputs to permitted commercial data and apply provider policies.
Check existing analyses and retrieved evidence before considering premium reasoning.
Use an allowed Pro Model only when the expected-value gate clears and issue a receipt.
The modelled ledger
What the same year of questions could cost
A thesis, not a case history. The assumptions are stated so you can replace them with your own numbers — which is exactly what a pilot does in week one.
Ledger rates are the vendors’ own published list prices: $0.0150 per premium question and $0.00065 per fast question at 1,500 input / 700 output tokens.
Modelled annual comparison
Workload: 100,000 questions/month for 12 months, at 1,500 input and 700 output tokens per question.
Published list prices, not estimates — premium baseline xAI Grok 4 at $3.00/$15.00 per 1M tokens = $0.0150/question.
Cheap metered tier xAI Grok 4 Fast at $0.20/$0.50 per 1M tokens = $0.00065/question.
Terminal-tier mix: 60% verified or in-perimeter, 30% Fast Models, 10% Pro Models. Rows exclude Qua fees, retrieval infrastructure, integration and human review.
WorkloadFrontier-onlyWith Qua
Verified answers and in-perimeter search720,000 annual questions resolved with no external model call.$10,800$0
Fast Models (Grok 4 Fast)360,000 annual questions at xAI's published Grok 4 Fast rate ($0.00065/question).$5,400$234
EV-gated Pro Models (Grok 4)120,000 annual questions keep frontier reasoning at the published Grok 4 rate.$1,800$1,800
At published vendor prices this thesis models $15,966 of avoided annual inference spend — $18,000 down to $2,034, a 88.7% reduction — before the excluded costs above.
Controls that matter here
Policy runs before routing, not after
Protect passenger data first
To support GDPR obligations, masking, provider blocks and maximum-tier clamps would run before routing, with Enterprise Search kept inside the perimeter. Qua Cloud, customer VPC or air-gapped deployment would be selected through the operator's security and privacy assessment.
Preserve approved maintenance authority
For US Part 121 operators, workflows would preserve the maintenance release requirements in 14 CFR 121.709 and point staff to controlled records. Qua would provide retrieval support, not certify airworthiness or replace authorised signatories.
Record information security decisions
For organisations in scope of EASA Part-IS under Regulations (EU) 2022/1645 and 2023/203, receipts would support review with exact inference cost, configured premium baseline cost, savings and what stayed private. Pro Models would require policy permission and an expected-value gate, while Open Models would be user-picked only; none of this constitutes aviation approval.
Week one would measure local-resolution rates and response times during a simulated disruption-driven question spike.
Week one would test document revision and fleet applicability, including whether superseded procedures are rejected.
Week one would compare receipt-level inference spend and approved-message accuracy against a matched premium baseline.
Sources
Every figure on this page, traceable
Market figures come from the publishers below. Qua savings are modelled from the vendors’ published list prices — they are not customer results.
[1]International Air Transport Association, Strong 2025 Passenger Demand Masks Ongoing Capacity Constraints (2026). Source Full-year 2025 passenger demand, measured in revenue passenger kilometres, increased 5.3%; the annual passenger load factor reached a record 83.6%.
[2]International Civil Aviation Organization, Annual Report 2023 — The World of Air Transport in 2023 (2024). Source Approximately 4.3 billion passengers were carried on scheduled air services in 2023; passenger volume does not measure AI-suitable operational questions.
[3]International Air Transport Association, Middle East Disruptions and High Fuel Prices Halve Airline Industry Profitability (2026). Source IATA's June 2026 outlook cut the 2026 industry net-profit forecast to $23.0 billion on higher fuel prices and Middle East disruption; these are forecasts, not final results.
[4]European Union, Regulation (EU) 2016/679 — General Data Protection Regulation (2016). Source GDPR requires lawful, purpose-limited and secure handling of covered passenger personal data, supporting access restrictions and privacy-by-design controls.
[5]European Union, Commission Regulation (EU) No 1321/2014 — Continuing airworthiness, Annex II (Part-145) (2014). Source EU Part-145 requires applicable current maintenance data and authorized certification of maintenance; generated answers do not replace those requirements.
[6]European Union, Commission Implementing Regulation (EU) 2023/203 — Information-security risks with a potential impact on aviation safety (2023). Source Covered aviation organizations must manage information-security risks with potential aviation-safety impacts and maintain prescribed records under the applicable Part-IS requirements.
