McKinsey’s Christian Langer, Joel Thibert, Dave Marcontell, Dan Leblanc, Joe Nutter, and Eric Porter explain why Gen AI tools will prove to boost productivity in a sector beset with labour shortages
Operating quietly behind the scenes, aircraft maintenance, repair, and overhaul (MRO) is one of the airline industry’s most essential functions.
Without it, airlines couldn’t achieve the remarkable feat of safely moving nearly five million people some 40 million air miles around the world each day.
Today, this indispensable industry is facing unprecedented headwinds. The recent double-digit growth of commercial air travel, a global shortage of aircraft, and a backlog of deferred maintenance from the Covid-19 pandemic are pushing demand for MRO services to new highs.
As airlines look to meet increasing passenger demand amid a constrained supply of new aircraft, the MRO industry will need to keep existing aircraft available, reliable, and in service for longer.
At the same time, the industry’s workforce has been squeezed and costs have soared. Driven by both inflation and a high demand for and low supply of workers, hourly rates for aircraft technicians and maintenance engineers have risen by more than 20 percent since 2019.
A wave of retirements has also led to a more junior, and thus less productive, workforce. In the upcoming years, these labour shortages are expected to persist and even intensify.
By 2033, one-fifth of the industry’s maintenance technician jobs are projected to go unfilled. Additionally, airlines and MRO players are also wrestling with supply chain disruptions and materials cost inflation.
Generative AI [Gen AI], which has advanced rapidly in just the past year, represents a promising capability to tackle some of these challenges.
The technology – which can generate relevant content from heaps of data in response to human prompts – is already re-shaping the future of work and transforming productivity across other industries.
Gen AI tools are particularly well-suited to knowledge-based and data-intensive businesses like aviation MRO.
Many roles within the aircraft MRO industry rely on the analysis and interpretation of a wide variety of differently formatted and sourced information, including manufacturer and operator service manuals, maintenance work orders, detailed descriptions of maintenance tasks (job cards), technician notes, and pilot writeups, as well as large volumes of aircraft sensor and instrument data. On average, today’s aircraft generate 50 times more input than older generations.1
To organise and interpret all this data, airlines and MRO providers are already taking advantage of a variety of broad-spectrum AI technologies, such as predictive analytics and machine learning.
The industry’s experimentation with large language models and Gen AI, however, is in its early stage and, to date, has mostly focused on applications that seek to enhance revenue, passenger engagement, and customer loyalty.
Last year, with boardrooms buzzing over the promise of Gen AI, airlines began dipping their toes in the water.
They created pilot projects for consumer-facing applications like call centres and trip planning, and used gen AI tools to help automate and trim the costs of software development. Use-cases for airline operations, however, were few and far between.
Most airline CFOs and CIOs were not yet convinced there were business cases to justify investing in new gen AI technology.
Today, the outlook is more auspicious. Many airlines and MROs see gen AI tools and solutions as a means to help both maintenance and back-office employees do their jobs more easily and efficiently, and with greater ingenuity.
Given the acute labour shortages in the MRO industry, these capabilities could turn out to be substantial productivity levers.
There is also reason to believe that gen AI platforms could boost the quality, consistency, and accuracy of maintenance work, ultimately keeping more aircraft in the sky and minimising aircraft out-of-service periods.
We see potential for airlines and MROs to use Gen AI in several ways:
Virtual AI maintenance and repair experts (“co-pilots”): On the shop floor, in hangars, and at line maintenance stations, aircraft technicians spend a substantial amount of time researching and trouble-shooting problems.
Gen AI could streamline this significantly by allowing workers to have a “conversation” with their data. Imagine a mechanic asking a digital assistant: “A compressor is leaking. What might be the issue?” or “Pull up the notes from my last shift.”
Workers would be coached through the identification of the most likely initial resolution, with a technically literate assistant providing recommendations on problem-solving and next steps.
Gen AI could generate this content from a relatively simple “reading” of unstructured and underutilised sources of information, including maintenance manuals and data from previous repairs.
One leading regional airline, for example, is currently building a pilot project where workers type a question or problem into a chat box, which uses gen AI to identify and provide the relevant section of a manual.
AI-augmented reliability engineering tools:
To ensure that aircraft equipment can function without failing while in-service, teams of reliability engineers sift through vast quantities of unstructured maintenance records.
The engineers look for failure patterns and assess whether the right aircraft maintenance tasks are happening. Gen AI tools can transform the productivity of this process by doing the sifting, pattern recognition, and analysis almost instantaneously.
This information could then be accessed in response to prompts from engineers. The reliability team at a leading US airline MRO, for instance, is experimenting with using such tools to extract failure patterns from maintenance logs and automatically set up planned maintenance tasks.
This gen AI augmentation reduces day-to-day toil for engineers, freeing up their time so they can use their expertise to solve the toughest reliability problems. Ultimately, this could lead to the proactive identification of fixes to reliability challenges, reducing aircraft out of service time.
In both this and the above-mentioned scenario, it is up to the engineer or technicians to make decisions about how to act upon the information given to them from gen AI models. Humans are always in the driver’s seat.
Assistants who take care of busy work: Once gen AI “co-pilots” have assisted with troubleshooting or repairs, they could be called upon to fill out reports that document this activity, saving workers significant time.
They could also be trained to automatically generate and submit work and purchase orders for needed replacement parts or additional servicing.
Instead of spending 60 percent of their day,2 researching, troubleshooting, and doing manual report preparation, technicians could spend more on “wrench time” – doing actual technical work.
In the back office, gen AI could generate similar efficiencies. Additional record-keeping automation and auditing would help procurement, HR, finance, and administration personnel identify gaps or inconsistencies in data and flag potential non-compliance issues.
Some airlines are looking at gen AI tools that could be prompted to integrate the records for newly acquired aircraft into their ERP systems. Currently, this effort involves weeks of manual review and migration of an aircraft’s maintenance history.
Permanent quality control supervisors:
In recent years, some airlines that have less-experienced technician workforces have seen increased rates of mistakes or missed items during maintenance, frequently leading to costly repairs or rework.
To address this problem, some airlines are considering using cameras to record maintenance work and detect any mistakes or skipped steps.
In this scenario, AI learning models would be deployed and tuned to analyse continuous video feeds, then flag incidents for managers to evaluate. Managers could then interact with a gen AI-enabled “supervisor” to create targeted feedback on a potential problem.
This concept is similar to airlines’ highly successful Flight Operations Quality Assurance (FOQA) programmes.
Instead of video, FOQA programmes use detailed positioning data to do a post-flight comparison of an aircraft’s actual position relative to what the flight plan and approach indicated it should be. Deviations and trends are analysed to improve pilot training and performance.
A supply chain manager:
Many companies operate some type of supply-chain control tower – a connected dashboard of data, key business metrics, and events across their supply chain.
Gen AI-enabled predictive analytics tools can augment this command centre by analysing disparate communication and delivery patterns to automatically identify early warning signs of any supply or delivery problems.
Supply chain analysts could use a gen AI chatbot to dig deeper into these problems or get recommended actions to mitigate them.
By helping employees to do their jobs more efficiently, these use cases hold promise for helping airlines and MRO players alleviate some of their workforce challenges.
Evidence from other industries suggests the potential for significant productivity improvement. One mining company, for instance, is projected to see a 35-70 percent reduction in time it will take technicians to troubleshoot equipment problems and a 25-50 percent reduction in time needed to do unplanned repairs, once its gen AI root-cause analysis and testing support tools are fully scaled.
Gen AI tools and learning models also represent useful avenues for skills training – accelerating the onboarding of new hires, supporting the continuous upskilling of existing employees, and helping ensure that institutional knowledge and expertise doesn’t walk out the door every time an employee retires.
It’s not just about the technology: Getting the basics right
Recent experience has shown that capturing gen AI’s potential value is harder than expected. Developing the technology itself, while no small feat, is just one side of the coin.
Some two decades ago, when advanced analytics use cases like predictive maintenance first emerged, many airlines faced challenges figuring out how these solutions could really drive value.
Companies, for instance, went through a prolonged period of trial and error before landing on predictive maintenance solutions that could successfully identify potential problems with aircraft and drive reductions in downtime.
Today, the challenge is similar. Airlines and MROs are grappling with how to move gen AI beyond provocative and toward profitable. To get on this pathway, airlines and MRO players will need to solve for several challenges.
Striking the right balance between careful and agile.
As a regulated, safety-dependent, and capital-intensive industry with historically low margins, airlines and MROs don’t have the luxury of being able to deploy the “fail fast” approach – throwing resources at many experiments at once to see what will work.
As a result, the industry has historically been slower to embrace change and implement cutting-edge technologies, especially in non-customer facing areas.
Many airlines and MROs are still in the process of migrating from legacy systems and use paper or pdf- based maintenance records in their interactions with third party vendors. Addressing this technology deficit and “dirty-fingerprint” recordkeeping is a significant hurdle to unlocking the potential of gen AI.
Pairing gen AI virtual experts with existing maintenance systems requires digitised processes, an integrated IT architecture, and available digital data.
Implementing this without impacting operations – or even more importantly, safety or airworthiness – will mean collaboration between numerous cross-functional stakeholders. Solutions must also be thoroughly tested and retested prior to deployment, as failure or repeated iterations are not viable options.
While gen AI is an enabler for more efficient operations, it isn’t a cure-all. The technology must be layered on top of effective, carefully-considered management strategies and ways of working to avoid roadblocks.
Preserving strict safety and regulatory compliance.
Maintenance of commercial aircraft is a high-stakes operation – safety is non-negotiable. If accuracy isn’t impeccable or quality fails to live up to a high standard, severe consequences can result.
In the near-term, gen AI platforms won’t be flawless – and thus airlines and MROs should not blindly rely on them in critical scenarios. To manage safety risks, the best AI use cases are those that accelerate and augment human judgement – a gen AI “co-pilot”.
These applications, however, will still need to be accompanied by significant investment in rigorous and stringent quality assurance and quality control processes. In order to address regulator concerns, humans will need to be trained to catch and remove potential gen AI “hallucinations,” or the generation of false or misleading information.
Finding the right talent. The skills needed to develop and integrate gen AI solutions go beyond technical acumen.
In addition to past experience building and tuning large language models, data scientists and software developers need a range of other abilities, such as design skills to understand where and how gen AI solutions should be focused and strong forensic skills to figure out the causes of breakdowns.
In addition, in order to understand the types of high-quality answers gen AI platforms will need to generate, some of this tech talent will also need airline maintenance expertise – a combination of skills not easy to find.
As an alternative, airline and MRO players can hire translators to help facilitate communication and collaboration between tech talent and frontline maintenance staff.
Having the right data.
The aviation industry generates large amounts of operational and performance data, but in order to use them for gen AI platforms, airlines and MROs will need to either own or have access to this wealth of data. This could necessitate partnerships between airlines, MRO suppliers, and manufacturers.
Immediate steps to get ready
Still recovering financially from the effects of the Covid-19 pandemic, many aviation players are reluctant to make large investments in new technology that does not yet have a demonstrated business case.
“[Gen AI] is still a technology with lots to explore in order to understand how well it will work or not. We have some challenges from that, especially in how we will monetise it,” says one airline operational vice president.
While measurable value is still potentially years away, companies can take action now to get started. These three steps will allow airlines and MRO providers to capture quick wins and be ready for a full integration of gen AI into their operations as new technological use cases appear:
Focus on top priorities.
Testing gen AI use cases is relatively easy. Getting these pilots to scale so they can drive meaningful value is difficult.
To find the use cases most relevant for a company’s challenges and priorities, leadership will need to first answer the question: Which opportunity is greater – workforce efficiency or quality control/risk reduction?
One aerospace company held an off-site ideation workshop to formulate an answer and identify where the major sources of value were within their organisation and within each domain. From there, they created a clear roadmap for the development of gen AI use cases, including the necessary enablers.
Though the precise prioritisation of gen AI use cases will vary by company, many airlines and MROs may find short-term opportunities in access to digital records, troubleshooting chatbot co-pilots, automation of compliance audits, and a virtual assistant for inventory planning.
Decide how and where to play.
It may sound obvious but without a clear strategic vision, companies can easily veer off in too many directions.
The best use cases for gen AI will likely vary between MRO providers and aircraft MROs are likely to find that early adoption will help give them a competitive advantage, reduce costs, and afford them the option of generating revenue by selling their solutions to airlines.
Airlines, on the other hand, may prioritise improving aircraft performance (eg reducing out-of-service time), and focus on their own operations instead of third-party monetisation. All types of players will need to decide whether to build, buy, or partner to develop the technology.
One large regional airline, for instance, decided that, in cases where AI software vendors tend to move slowly when customising for specific customers, the company would develop its own products designed to fit its particular work locations and strategy.
Move to implement now.
With Gen AI technology rapidly evolving, it might be tempting to wait and see how things play out. We believe this would be a mistake. Companies seeking to gain a competitive advantage will want to take immediate action across three dimensions:
Building the underlying infrastructure to enable Gen AI. In addition to building all the layers of the Gen AI tech stack (eg cloud computing) and doing technical talent acquisition and/or upskilling, companies will need a clear data strategy that includes a path to gain access to data via partnerships or ecosystems.
This can then be used to develop a structured and adaptable data infrastructure that makes multiple sources of information available for Gen AI solutions to use.
Creating quick wins in areas with fewer regulatory. Early use cases that can help airlines and MROs create internal enthusiasm for Gen AI are those that build upon existing capabilities.
For example, adding intelligent, natural language queries to existing digital search functions for maintenance records, manuals, and job cards could demystify gen AI and quickly demonstrate value by enabling quick productivity wins across maintenance functions.
Another early win is the addition of Gen AI capabilities to current predictive maintenance analytics. Augmenting existing machine learning capabilities with Gen AI, which does a better job of leveraging unstructured data, can improve forecasting accuracy and allow airlines to better plan for the unplanned.
Gen AI solutions can better process and incorporate “human data,” such as pilot writeups into predictive models, further improving performance.
Laying the groundwork for deeper integration. Longer-term, airlines and MROs will explore use cases beyond adding enhancements to existing capabilities.
Instead, companies will aim to embed the intelligence of Gen AI across the airline maintenance value chain. Doing so will introduce fundamentally different ways of working.
The frontline workforce, for example, will not just use AI-enabled digital assistants, but develop a new set of skills – continuously understanding and monitoring risks, then proactively identifying and resolving issues before they impact operations.
Driving acceptance of this deeper engagement with technology, as well as comfort levels with these new skills, will require active engagement from leadership and a careful change management plan.
In Conclusion
In the face of any transformative technological change, leaders face critical choices. Within a conservative industry like aviation MRO, it can be tempting to minimise risks by taking a “wait and see” approach and jumping in later.
But this stance also comes with substantial risk. At some point in the near future, Gen AI solutions will fundamentally disrupt the airline maintenance industry – sitting on the sidelines won’t be an option.
Players that engage now to develop a clear strategic vision with investment in value-focused experimentation will be better positioned to realize the value of Gen AI.
About the Authors
Christian Langer is a partner in McKinsey’s Hamburg office, Joel Thibert is an associate partner in the Calgary office, Dave Marcontell is a partner in the Atlanta office, where Joe Nutter is a consultant, Dan Leblanc is a partner in the Dallas office, and Eric Porter is a consultant in the Boston office.
- https://www.lufthansa-technik.com/en/aviatar
- McKinsey research
