AI in aerospace: Scaling from pilot to production

Artificial intelligence in aerospace applies machine learning, computer vision and automated decision-making to the design, production, operation and sustainment of aircraft and the defense systems around them. Its most valuable aerospace and defense applications are increasingly forms of operational AI (systems that act on what they find), spanning aerospace manufacturing automation, predictive maintenance and autonomous aircraft technology, with human-in-the-loop AI keeping people accountable for the critical calls. Used well, these capabilities help organizations work around workforce constraints, improve production efficiency, increase aircraft availability and finally move AI initiatives from pilot to production.

For executives, the tension is hard to ignore. Demand has rarely been stronger, while the workforce and supply chain built to meet it are under real strain. Artificial intelligence in the aerospace industry offers a way to get more from the people and assets already in place, provided it makes the leap from promising pilot to dependable production.

The organizations making that leap tend to treat AI as an operating capability instead of a series of experiments. This blog looks at how that plays out across the factory, the airspace and the path to autonomy, drawing on DXC's work in aerospace and in comparably demanding, regulated fields to show what it takes to scale trustworthy AI.

The squeeze is structural, not cyclical

Start with demand. Boeing's Commercial Market Outlook¹ projects a need for roughly 43,600 new commercial aircraft over the next two decades, with the global fleet nearly doubling by 2044. That's a generational build-out.

Output hasn't kept pace. Aircraft supply remains well below pre-COVID levels, and recent delivery volumes resemble those of the early 2010s despite record air travel demand. Defense adds pressure of its own, as governments accelerate rearmament and modernization in response to rising geopolitical tensions and compressed timelines.

The workforce is the primary constraint

An Aerospace Industries Association workforce study² found that 76% of companies struggle to hire engineers and 56% face skilled-trade shortages, with attrition near 15% (more than double the U.S. industry average). Nearly a third of manufacturing and engineering roles are held by people over 55, so the knowledge gap widens with every retirement. Add a fragmented supply chain that still depends on manual planning, and the scale of the problem becomes clear.

The industry can't hire its way out of this. It has to get more from the people and assets it already has, which is exactly what AI, robotics and automation are designed to do.

That inevitably raises two questions in design offices and on shop floors. Will AI take over aerospace engineering? And if it won't, how will AI affect aerospace engineering in practice? So far, the relationship between AI and aerospace engineering looks more like a partnership. AI for aerospace engineering takes on repetitive analysis, inspection and planning work, which frees experienced engineers to focus on judgment calls and to pass on what they know before they retire.

What’s changed for AI in the aerospace industry?

That's the uncomfortable truth about AI in the aerospace industry: buying into it is easy, but making it pay off is anything but. Gartner predicts³ that organizations will abandon 60% of AI projects unsupported by AI-ready data through 2026, and its research shows that most organizations aren't confident their data practices can support AI at all. MIT's Project NANDA⁴ found that only about 5% of enterprise GenAI pilots produced measurable value, with the rest stopping short of any viable return.

This doesn’t mean AI is underperforming

Analytical AI (the kind that predicts and recommends) is now standard practice, and today's tools reliably forecast component failures and flag supplier risks. The value still on the table is harder to land, and it depends more on the surrounding operating model than on the AI model itself. Buying more aerospace AI software won't close that gap on its own.

Success comes from AI-ready data, disciplined governance, realistic testing and a clear line from each use case to a business outcome. In aerospace, that's harder than it sounds, because the relevant data is scattered across engineering, production and sustainment systems that were never designed to talk to each other. The hard part now is industrializing AI so it holds up in daily operation, and that's the shift aerospace leaders need to make.

Identifying value across AI applications in aerospace

Not all AI applications in aerospace carry the same risks and rewards, and it helps to separate them into two kinds:

• Analytical AI predicts. It forecasts part failures, flags quality defects and spots anomalies in supplier invoices. The technology is well established, and most of the AI and ML in aerospace in use today falls into this category.

• Operational AI acts. It manages factory processes, guides inspections, helps direct air traffic and flies aircraft. This is where the biggest opportunities in aerospace lie, and it demands the highest level of trust because errors have real consequences.

The distinction matters for investment. Analytical AI supports human decision-making, while operational AI takes on at least part of the decision itself, which raises the bar on accuracy, security and accountability. That's why the operational frontier has developed more slowly. Building these systems safely takes discipline and a strong supporting ecosystem that few organizations have assembled, and those that manage it will pull well ahead.

Operational AI is already paying off

The encouraging news is that operational AI in aerospace is no longer theoretical. It's already producing measurable results, and three domains show where the value concentrates.

The first is the factory and its supply chain, the source of many of the industry's cost overruns and delays over the past two decades. Automotive digitized and automated early, helped by a relatively stable base of large manufacturers that could co-invest. Aerospace has lacked that volume and coordination, so much of its production has stayed manual, with quality and schedule hinging on human planning.

The practical answer is general-purpose robotics, particularly collaborative robots (cobots) that work safely alongside people on the line. Their advantage is flexibility. Traditional automation has to be custom-programmed task by task, which rarely pencils out at aerospace volumes. AI lets a cobot learn a wide range of tasks without expensive retooling, and that finally makes AI automation in aerospace affordable at industrial scale.

The outcomes are already visible in aerospace manufacturing

Using DXC Smart Factory solutions⁵, one aerospace manufacturer increased throughput twelvefold with the same equipment and labor footprint, saving roughly $15 million a year. Edge AI now inspects parts in real time on the line, catching defects earlier and lifting first-pass yield. For AI in aerospace manufacturing, that matters enormously, because a single missed flaw can ground an aircraft.

Underneath it all, a connected digital thread tracks and optimizes machine usage and labor automatically, bringing the conformity and control the industry has long needed.

The same intelligence extends beyond the factory gate. Predictive maintenance in aerospace reads sensor data across a fleet to fix parts before they fail, increasing aircraft availability and trimming the aftermarket costs that consume a large share of an aircraft's lifetime spend.

The second domain is the airspace itself, and it's where much of the future of AI in aviation will play out. Drones and air taxis are moving from demonstration to service, and the numbers are striking. The advanced air mobility market⁶ is projected to grow from about $11.75 billion in 2024 to roughly $137 billion by 2035. That implies a sharp rise in the volume and variety of low-altitude traffic, much of it below 3,000 feet, where little structured management exists.

Human controllers can't scale to that complexity, and giving more of them ever smaller slices of sky won't solve it either. Managing this safely will require AI-assisted air traffic management that can evaluate and coordinate far more movement than any human team, while keeping people in command of the decisions that matter most. The same logic applies to the growing fleet of uncrewed cargo and surveillance aircraft that defense operators now field.

The third domain is autonomy, the most ambitious application of AI in aerospace. It's worth being precise here, because autonomous and unmanned mean different things. Most drones are unmanned but still remotely piloted from a ground station. True autonomy means an AI system continuously monitors the airspace, makes decisions and executes its mission without human control, and the case for autonomous aircraft technology closely mirrors the case for autonomy on the ground.

Deployed correctly, autonomy delivers greater efficiency and a genuinely safer system, because it removes the fatigue and inconsistency behind so many incidents. The condition is that the system earns that trust before it's ever allowed to fly.

Clearing that bar is where the hardest and least glamorous work begins.

What trustworthy operational AI really takes

Every board should be asking a simple question at this point: what are the risks of AI in aerospace, and how do you contain them? In operational systems, the biggest risks sit in the data and the testing, and many programs underestimate the effort both require. Building a certifiable operational AI system takes much more than a strong algorithm. It needs data of a scale and quality far beyond typical analytics projects, managed diligently throughout the development life cycle.

A practical way to think about that data is in tiers:

1. Build a dataset grounded in the underlying physics of the system being modeled.
2. Develop a second dataset that reflects the real operating environment, including its complexities and edge cases.
3. Use algorithms to generate synthetic data, expanding the set significantly and covering scenarios that are too rare or hazardous to capture in real-world conditions.

Each tier has to be hosted, curated and governed with rigor, because the quality of the training data ultimately determines how the system performs.

Testing is just as demanding

A robust simulation environment that reflects real conditions is often the deciding factor in acceptance and, where required, certification. Certifying an airborne system is fundamentally about proving it's safe, which makes simulation at scale a core discipline of AI in aerospace engineering. You can't prove safety without exposing the AI to millions of realistic situations long before it leaves the ground. Add human-in-the-loop oversight and clear governance, and you have the makings of AI-driven aerospace systems worth trusting.

This is the discipline DXC has built into Xponential⁷, a repeatable blueprint for moving AI from a credible pilot to dependable production, designed for regulated environments where the cost of error is high. Running these systems reliably at scale brings its own challenges. DXC OASIS⁸ layers a single, governed view across complex technology estates, pairing human insight with AI agents. Teams can then focus on critical issues instead of chasing alerts, and isolated wins become a repeatable capability.

Proving AI can scale in high-stakes environments

The most persuasive evidence comes from settings where a wrong answer carries real consequences. DXC worked with Singapore General Hospital⁹ to develop an augmented intelligence solution that guides antibiotic decisions for lower respiratory tract infections. With clinicians in the loop, the system achieved 90% accuracy, improved clinical outcomes and helped reduce antimicrobial resistance.

What carries over to aerospace is the pattern. Both fields depend on regulated, safety-critical, human-in-the-loop systems supported by disciplined data management and governance.

The pattern repeats closer to home

The European Space Agency¹⁰ worked with DXC on a GenAI platform that lets its people sift and examine vast volumes of research data far faster than before. In defense, where data sovereignty and security are non-negotiable, DXC has developed capabilities such as Secure Sovereign AI and VisionIQ to support private AI for aerospace and defense programs inside tightly regulated environments.

That mix of industry depth and delivered outcomes is one reason Avasant named DXC a leader¹¹ in its 2025-2026 Aerospace and Defense Digital Services assessment.

Making operational AI a core infrastructure

The organizations pulling ahead treat AI as core infrastructure and fund, staff and govern it like any other mission-critical system. For leaders shaping an aerospace AI strategy, five priorities follow:

• Start with high-value operational use cases where the payoff is measurable, such as factory throughput, quality inspection or predictive maintenance, and prove the model there before scaling it.

• Strengthen data foundations first, because inadequate data is where many AI projects come undone.

• Invest in the foundations of scale (simulation, testing and governance) on a shared AI aerospace engineering platform, so each new use case doesn't start from scratch.

• Build in human-in-the-loop oversight from day one to earn trust and support compliance.

• Choose a partner with a proven blueprint for taking AI solutions for aerospace from pilot to production, because stalled projects delay programs and cede ground to competitors.

The future of AI in aviation

Demand is there, the workforce gap won't close on its own and the technology is mature. AI in aerospace and defense will reward the organizations that industrialize operational AI now and do it with the discipline safety-critical work demands. Every quarter spent in pilot mode is capacity the industry can't afford to lose, so the time to choose your first production use case, fix its data and set its governance is this planning cycle.

Connect the dots

If you're exploring where operational AI could add the most value for your organization, our aerospace and defense team is always ready to compare notes. 

References 

¹ Boeing, Commercial Market Outlook. https://www.boeing.com/commercial/market/commercial-market-outlook

² Aerospace Industries Association, AIA and McKinsey study on tackling talent gaps in aerospace and defense. https://www.aia-aerospace.org/news/aia-and-mckinsey-release-new-study-on-tackling-talent-gaps-in-aerospace-and-defense-industry/

³ Gartner, Lack of AI-ready data puts AI projects at risk. https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk

⁴ MIT Project NANDA, The GenAI divide: state of AI in business 2025 (reported by Fortune). https://finance.yahoo.com/news/mit-report-95-generative-ai-105412686.html

⁵ DXC Technology, Aerospace and defense solutions. https://dxc.com/industries/aerospace-and-defense

⁶ Grand View Research, Advanced air mobility market report. https://www.grandviewresearch.com/industry-analysis/advanced-air-mobility-aam-market-report

⁷ DXC Technology, DXC Xponential. https://dxc.com/insights/ai/xponential

⁸ DXC Technology, DXC OASIS. https://dxc.com/platforms/dxc-oasis

⁹ DXC Technology, DXC unveils Xponential (Singapore General Hospital). https://dxc.com/newsroom/10282025-dxc-unveils-xponential-a-repeatable-blueprint-to-accelerate-enterprise-ai-outcomes

¹⁰ DXC Technology, Generative AI (European Space Agency). https://dxc.com/solutions/ai-data/generative-ai

¹¹ DXC Technology, DXC named a leader in Avasant Aerospace and Defense Digital Services assessment. https://dxc.com/newsroom/07062026-dxc-named-a-leader-in-avasant-aerospace-and-defense-digital-services-2025-2026-radarview