AI Travel Aviation. 95% Zero-Touch. 3 Tariff Worlds. One Agent.
Cockpit, Cabin, Ground. IROP Volatility. International Per Diems.
The AI Agent classifies crew group, country per diem, and tariff application per rotation. The rule engine calculates reimbursement amounts and tax allowances deterministically. Humans stay where tariff interpretation or tax law requires it.
Cases/year (simulation)
Zero-touch rate
IROP rate (industry standard)
Highest rule complexity
Cases and zero-touch: Gosign simulation model. IROP rate: EUROCONTROL / US DOT BTS, 2024. (UK: UK CAA statistics post-Brexit.)
What the Agent classifies
Three crew groups, one agent
Cockpit, cabin, ground - each group with its own tariff, per diems, complexity. You know the complexity. Here's how the AI Agent solves it:
Hundreds of rotations per day
Every rotation creates a travel expense case. The AI Agent reads rotation data from crew planning systems and triggers per diem calculation automatically. No manual capture, no form, no delay - hundreds of cases processed per hour without human input.
Three to five countries per duty day
A single rotation touches multiple countries in one duty day. The AI Agent classifies each country segment and its duration. The rule engine then selects the correct per diem rate per country and calculates full or reduced amounts based on stay duration - deterministically, for every rotation, without exception.
Tariff override per crew group
Cockpit, cabin, ground and maintenance each have different collective agreements. The AI Agent classifies crew group per employee. The rule engine then applies the correct tariff override - collective agreement rates that supersede statutory rates, company agreements that add a further rule layer. Same rotation, different crew group, different calculation - automatically.
IROP volatility in real time
Delays, diversions, repositioning change every rotation. The AI Agent classifies the type of irregularity and maps the actual itinerary. The rule engine recalculates per diems based on actual location and duration - not the planned schedule. Volatile operations, deterministic results.
Volume governance without sampling
Thousands of crew generate identical cases daily. The AI Agent checks 100% of all cases against the same rule version. Every systematic error is caught immediately - not discovered during quarterly reconciliation. Tax audit risk eliminated at the source, not managed after the fact.
One rotation. 40 to 120 micro-decisions.
A crew member flies a three-day rotation with layover. The Decision Layer decomposes this into individual decision steps:
| Step | Decision | Decider | Rationale |
|---|---|---|---|
| 1 | Read rotation from duty roster | Automatic | Data import, no decision |
| 2 | Determine country sequence | Rule engine | GPS or schedule: which countries, which duration |
| 3 | Select per diem per country | Rule engine | Country-specific rates per regulatory framework |
| 4 | Check collective agreement override | Rule engine | Crew-specific collective agreement overrides statutory rates |
| 5 | Calculate meal deduction | Rule engine | Provided meals reduce per diem deterministically |
| 6 | Classify IROP | AI | Delay, diversion, repositioning: AI classifies type |
| 7 | Recalculate IROP impact | Rule engine | Changed duration or country: recalculate per diem |
| 8 | Check layover hotel | AI + Rule engine | AI extracts hotel data, rule engine checks policy compliance |
| 9 | Map cost centre | Rule engine | Rotation to fleet, fleet to cost centre |
| 10 | Generate audit pack | Automatic | Sealed decision record per rotation |
Simulation
Calculated for aviation volumes
We configured the Travel Decision Layer with realistic aviation parameters and ran the numbers. The results show what changes at enterprise volumes.
Simulation parameters
| Crew members | 10,000 to 50,000+ (multiple staff groups with separate collective agreements) |
| Collective agreements | 2 to 5 in parallel (per staff group and carrier) |
| Cases/year | 100,000 to 1,000,000+ |
| Jurisdictions | Multi-jurisdiction (EU member states, UK, US and beyond) |
| System integration | Crew planning → Decision Layer → ERP/Payroll |
| IROP rate | 10 - 20% of all rotations with schedule deviation (industry standard) |
Before / After
| Dimension | Manual | Decision Layer |
|---|---|---|
| Processing cost | from EUR 53 (GBTA)* | < EUR 9 |
| Error rate | 19% (GBTA) | < 0.3% |
| Processing time | 5 - 12 business days | Minutes |
| Zero-touch rate | 0% | 95% |
| Audit readiness | Manually reconstructed | Automatically generated |
| Collective agreement change | Weeks | < 24h |
| Retro correction | Manual overwrite | Reversal + adjustment (append-only) |
* GBTA Foundation 2024: USD 58 per transaction (approx. EUR 53). Crew cases with collective agreement logic, multi-jurisdiction and IROP handling are typically higher. IROP rate: EUROCONTROL Annual Report, 2024. Simulation results: Gosign model calculation.
In the simulation, the Decision Layer achieves a zero-touch rate of 95% - regardless of whether 100,000 or 1,000,000 cases per year are processed. Only 5% require human attention. For the remaining 95%, a complete, audit-ready decision record is available. Corrections and collective agreement changes generate reversal and adjustment entries - no overwrites, complete posting history.
Architecture and implementation
The Travel Decision Layer runs entirely within your infrastructure: your data centre, your network, your control. No SaaS dependency, no data leakage. For aviation, this means: crew planning system integration, real-time rotation data processing, collective agreement-specific rulesets per staff group and end-to-end audit trail to SAP FI/CO. Typical pilot projects start within 3 months with one crew group and one collective agreement.
Deep Dive in the Agent Briefing
Our article series for decision-makers implementing AI agents in the enterprise.
Travel Decision Layer in Other Industries
Each industry has its own tariffs, cost structures, and complexity drivers. The Decision Layer is the same. The configuration is industry-specific.
Frequently asked questions about the aviation configuration
Can the Decision Layer handle multiple collective agreements simultaneously?
Yes. Each staff group (cockpit, cabin, ground, maintenance) can have its own collective agreement ruleset. The Decision Layer selects the correct ruleset per employee per rotation based on staff group classification.
How are IROPs (irregular operations) handled?
AI classifies the type of irregularity (delay, diversion, repositioning). The rule engine then recalculates the tax implications based on the actual itinerary rather than the planned rotation. The audit trail documents both the original plan and the actual execution.
What does renegotiation of collective agreements mean for the system?
Decision tables are versioned. New collective agreements are deployed as their own version with a validity period. Retroactive changes trigger automatic recalculation of all affected cases - no entries are overwritten. The Decision Layer generates reversal and adjustment entries. The complete history is preserved. In the simulation, the changeover takes less than 24 hours.
Does the system integrate with crew planning tools?
The Decision Layer is configurable to read rotation data from standard crew planning systems. The integration maps duty rosters to tax-relevant events without requiring crew members to enter data manually.
How does a project start?
In a 30-minute call, we clarify your parameters: crew volumes, collective agreements, system landscape. This produces a pilot proposal: one staff group, one collective agreement, one data flow. Typical pilot projects start within 3 months.
Let's do the maths.
30 minutes. Your crew volumes, your collective agreements, your results. We configure the Decision Layer with your real parameters.
Book a Call