A summer thunderstorm is moving toward a busy European airport. Dozens of flights are at risk of being canceled or delayed. Crews are close to their legal limit on hours worked. Aircraft are parked in the wrong places for tomorrow's first departures.
AI can already predict the weather and model a response. The hard part comes next. Airline crews have to be rebooked, aircraft must be repositioned, and maintenance has to be rescheduled.
Ryanair is convinced AI can help with more of those decisions. The airline runs more than 3,600 flights per day from 95 bases with almost 650 aircraft, and it has 300 more Boeing 737s on order. A single scheduling error can not only ground a plane, but also trigger a cascade of delays throughout the network.
In August, Ryanair announced a five-year partnership with Google Cloud. The airline will give its 35,000 employees access to Gemini Enterprise, Google's agentic AI platform, to help automate decisions and manage flight crew logistics.
Ryanair also plans to use two Google DeepMind models to assist with fleet operations and maintenance scheduling. Ryanair DAC CEO Eddie Wilson has tied the deal to a goal of 300 million passengers each year by 2034, up from around 216 million today.
Most airlines already use AI in operations control: 63% use the technology to manage disruption, aircraft assignment, and crew availability at once, according to SITA's 2025 Air Transport IT Insights report. Usage is lower for decision-making that requires consistent data from several sources. For example, 17% of airlines use AI to monitor turnarounds, the work of getting an aircraft ready for its next flight after it lands, in real time.
In a storm, the flight, crew, aircraft, and passenger systems need a single, current set of facts. Nearly half (49%) of airlines report data integration as the main barrier to that, SITA found.
The problem runs well beyond aviation. The ServiceNow Enterprise AI Maturity Index 2026, a survey of 4,500 executives in 19 countries and 12 industries, found that 71% of organizations struggle with the accuracy, access, and management of their data.
What would it take for an airline to trust AI to coordinate the response to a storm across 3,600 flights every day?
The deal brings Ryanair closer to physical AI—AI that can perceive, reason about, and interact with the physical world in real time. Some of the best-known examples are machines such as Waymo robotaxis and Amazon's warehouse robots.
The category also covers digital twins and sensor-based systems that help human teams manage complex physical environments. Ryanair's deployment is closer to this second bucket. Its AI agents generate recommendations, and people decide which aircraft fly which routes and when they go in for service.
Ryanair isn't the only company pushing AI toward physical operations. In April, Accenture invested in General Robotics, which makes software that connects robots from different manufacturers to AI models.
In January, PepsiCo announced a collaboration with Siemens and NVIDIA to build digital twins of its plants and warehouses, where AI agents can test changes before anyone alters the physical site.
Warehouse automation is further along. In June, Ambi Robotics and Pickle Robot announced an integration that automates logistics from trailer unloading to pallet stacking.
Unloading a truck is a fairly contained, repeatable task. A storm day is neither. Routing an aircraft around weather, or deciding when a jet needs unscheduled maintenance, involves a series of urgent, high-stakes judgment calls.
Even the narrower tasks are difficult to scale. In a survey of 300 manufacturing executives in North America and Europe, Tata Consultancy Services (TCS) found 77% expect physical AI to have a “significant or transformational impact in warehouse operations.” But 68% say they’re still experimenting with AI or haven't deployed it at all. TCS lists legacy system integration, data infrastructure, and workforce skills as the main hurdles.
Airlines run into these problems as well. Beyond the data barrier SITA found, BCG says legacy passenger service, enterprise resource planning, and maintenance systems make digital integration challenging.
Trusting AI to make airline decisions in a storm requires three things.
- A shared, current record of the facts systems have in common: When the scheduling system and the maintenance system both reference the same aircraft, they should draw on the same status, with the same definitions and the same time stamp. Without that, an AI agent can act with confidence on information another system has superseded. Across industries, only 16% of organizations have widely or fully replaced fragmented legacy systems with an integrated platform, according to the ServiceNow Enterprise AI Maturity Index.
- Orchestration, the rules that decide which system acts, in what order, and under what limits: Ryanair's plan combines a dual-cloud strategy, Gemini Enterprise, and two DeepMind models. Each needs rules for what it can do and when.
- Governance: An organization needs a record of what every AI agent did and why it acted. According to the Enterprise AI Maturity Index, the majority (61%) of top-performing organizations, which we call Pacesetters, installed a system to manage and track governance and compliance initiatives, compared to just 17% of others.
Many industries share the same problem as the airline sector: a scarce, expensive resource that has to be routed and maintained under constraints that shift by the hour. As Ryanair's bet plays out, other organizations will be taking notes.
Find out how ServiceNow can help you connect silos to get work done.
This article was prepared by ServiceNow without the involvement, review, or approval of any company discussed in it. Company names and trademarks are used for identification purposes only and do not imply sponsorship, endorsement, or affiliation.
