SITA OptiFlight Explained: The AI Tool Helping Airlines Reduce Fuel Costs and Emissions

The Problem With Generic Flight Planning

Most flight planning tools work from a fleet-wide performance model. They assume your A320 flies like every other A320. In practice, individual airframes accumulate wear, quirks, and performance variations that generic models simply miss.

SITA’s argument is straightforward: if you’re optimising for the average aircraft, you’re leaving real savings on the table for every aircraft that isn’t average—which is most of them.

How OptiFlight Actually Works

OptiFlight builds a digital twin of each individual aircraft. It pulls from three data sources:

• Quick Access Recorder (QAR) data — actual flight performance history from that specific airframe
• Machine learning models — trained on 40 flight parameters to understand how the aircraft behaves
• High-resolution 4D weather forecasts — real atmospheric conditions, not averages

The result is a set of flight-specific recommendations covering climb speeds, cruise altitude profiles, and descent paths—tailored to that aircraft, on that route, in those conditions.

Recommendations reach pilots via ACARS or an Electronic Flight Bag (EFB) interface. No new hardware. No cockpit modifications. After each flight, the system compares predicted versus actual fuel consumption and updates the model. It gets sharper over time.

Where the Savings Come From

The fuel reductions are incremental, which is exactly the point. Small improvements across thousands of flights compound quickly.

SITA’s reported figures suggest:

• Up to 5% reduction in climb fuel burn
• Around 3% savings during cruise
• Roughly 3% average savings across a full flight

That works out to approximately 73 kg saved per flight. Modest in isolation. At fleet scale, it becomes a material financial position.

The Sustainability Angle Is More Than Marketing

Airlines are under increasing regulatory pressure to prove their environmental claims. The EU Green Claims Directive, for instance, requires verified data—not aspirational targets.

OptiFlight generates a fuel consumption prediction before departure, then compares it against actual post-flight data. That creates an auditable trail. For airlines trying to substantiate CO2 reduction claims without being accused of greenwashing, that paper trail matters as much as the savings themselves.

ITA Airways, for example, rolled out OptiFlight across its fleet expecting to save more than 7,100 tonnes of fuel and reduce CO2 emissions by over 22,100 tonnes across 2025 and 2026. Those are specific, trackable commitments—not vague pledges.

OptiFlight Isn’t the Only Player

SITA isn’t operating in a vacuum. Airbus has its own Descent Profile Optimisation (DPO) tool, which focuses specifically on the descent and approach phase—one of the most fuel-intensive parts of any flight. Delta Air Lines deployed DPO across 270 aircraft in its Airbus fleet.

The two tools aren’t necessarily competing. Climb optimisation and descent optimisation address different phases of flight, and airlines with complex fleets may find value in layering multiple tools.

What’s notable is that both approaches share the same logic: the biggest near-term gains in aviation fuel efficiency aren’t coming from new engines or next-generation aircraft. They’re coming from flying existing aircraft smarter.

Why This Matters Now

New aircraft technologies take decades to reach meaningful fleet penetration. Sustainable aviation fuel supply remains constrained. The industry has committed to Net Zero 2050, but the path there requires solutions that work today, on current fleets, within existing operations.

AI-powered flight optimisation sits in that gap. It doesn’t require a capital expenditure cycle or a regulatory approval process. It requires data, a software integration, and pilots willing to follow a recommendation.

The Practical Takeaway

If you’re evaluating AI tools for operational efficiency in aviation—or studying how AI creates measurable ROI in asset-heavy industries—OptiFlight is a useful case study in what “applied AI” actually looks like in practice.

It’s not a chatbot. It’s not a dashboard full of insights nobody acts on. It’s a system that ingests real operational data, builds an individual model for each asset, and delivers specific recommendations at the moment they’re useful. The savings are auditable. The deployment doesn’t require hardware changes. And the business case compounds with fleet size.

That’s a pattern worth recognising, whether you’re in aviation or not.