Aerospace is hunting for talent it already has
The industry found 27,500 asteroids in data it had already collected. It can find its workforce the same way.
Article By: Emil Barr
In 2022, a graduate researcher named Joachim Moeyens ran a new algorithm across a month of old telescope images and found 104 asteroids nobody knew were there. The pictures had been taken years earlier, for other purposes, by instruments never pointed at the sky with asteroid-hunting in mind.
Nobody built a new telescope or launched a new mission. Somebody simply wrote software that could recognise an orbit in data that had been sitting in an archive since 2012.
Two years later, they ran the algorithm across the entire catalogue and came back with 27,500 high-confidence asteroid candidates in a matter of weeks. The project analysed 5.4 billion observations, and more than a hundred of the candidates were near-Earth asteroids. As the Asteroid Institute’s executive director put it, any telescope with an archive can now become an asteroid search telescope.
The algorithm is called THOR, and it runs on a platform built by the Asteroid Institute, a programme of the B612 Foundation, which I help fund, alongside billionaires and Queen’s Sir Brian May. I’m a 23 year old multimillionaire, and I put money into planetary defence for the unglamorous reason that a solvable catastrophe deserves solving. I run a workforce development company, which means I spend most of my time on a version of the same problem: what happens when the technology moves faster than the people trained to use it?
Aerospace is about to experience that gap at a scale almost no other sector will face, and its existing systems are increasingly poor at recognising what the people it already has can do.
The 2025 AIA and McKinsey workforce study found industry attrition stuck at nearly 15 percent, with 76 percent of AIA member organisations reporting sustained difficulty hiring engineering talent and 56 percent reporting difficulty sourcing skilled trades. That is against a workforce that had grown to 2.23 million employees, up 2.9 percent in a year, with 25% of them at more than twenty years of experience and at or beyond retirement eligibility. Meanwhile the supply chain is expected to stay under pressure through at least 2027, with shortages of materials and skilled labour among the causes.
That is one curve. Here is the other.
Deloitte’s 2026 outlook projects the share of industry job postings requiring data analysis skills rising from 9 percent in 2025 to nearly 14 percent by 2028, with data science, data engineering, AI, machine learning and statistical analysis among the fastest-growing skills in the sector. In the UK, Skills England and the Ministry of Defence project the fourteen priority occupations in defence growing by 53,000 workers between 2025 and 2035, a 58 percent increase, with a further 29,000 needed to replace people leaving. That is 82,000 positions to fill over a decade in one sector.
The British picture carries a warning that the American numbers obscure. Twelve of those fourteen priority occupations overlap with at least one other priority sector, and defence shares nine of them with advanced manufacturing alone. The same engineers, the same digital and technical roles, are simultaneously being counted in the workforce plans for other growth sectors — each of those sectors is therefore drawing on the same underlying pool of technical talent. Skills England also notes that defence’s priority occupations demand higher than average proficiency in digital literacy, problem solving and decision-making, which is precisely the profile every other growth sector says it needs.
Then there is the commercial side. Boeing’s 2026 Pilot and Technician Outlook projects global demand for more than 2.4 million new aviation professionals through 2045, including 728,000 maintenance technicians. Two-thirds of that demand is replacement of people who are leaving, not growth.
This means that the requirement is rising sharply at precisely the moment the supply is ageing out, and the competition for the remainder is intensifying across every adjacent industry. Companies have responded the way most industries respond: wider recruiting channels, higher wages, better onboarding, more places to look. Those are reasonable moves, but they do not solve a problem created partly by the way the industry defines qualified talent.
When an industry pulls harder on the same lever for years and the number doesn’t move, the lever probably isn’t connected to the machine — the framing is what’s broken.
A shortage implies scarcity of supply, and scarcity implies a population of qualified aerospace professionals somewhere out there who simply haven’t been contacted. That population does not exist at anything like the scale the industry needs, but what does exist is a much larger population of people who could do a meaningful share of this work within a year, and a hiring apparatus with no mechanism for recognising them.
Aerospace has been organised for decades around deep specialisation, and those disciplines are not disappearing. They are converging. An engineer may now need simulation, digital twins and enough machine learning to interrogate a model alongside the fundamentals of aerodynamics. A maintenance technician increasingly works with predictive systems that flag a failure before an aircraft is grounded. A defence analyst can process in seconds what used to take a week, and remains entirely responsible for deciding whether the output is sensible.
Deloitte’s read is that the emphasis in the sector is shifting away from hiring AI specialists and toward embedding AI fluency across the entire workforce. That is a different problem from recruitment, and it has a different solution.
The best aerospace engineer of the next decade will probably not be the person who knows the most about AI, but rather the engineer who knows enough about AI to use it well while carrying twenty years of accumulated knowledge about how aircraft actually behave. Those are not necessarily new people, but current people with an added capability, and for many roles that capability can be developed through targeted training rather than another degree.
There is a specific obstacle worth naming, because it is not the one most executives expect: Deloitte’s outlook observes that while senior leaders are broadly optimistic about AI, middle management is often more sceptical, untrained and risk-averse. That is the layer where a training programme either reaches the floor or quietly dies, and it is rarely the layer these strategies are designed for.
The AI talent question in aerospace presents as a recruitment problem and behaves as a training problem. You cannot recruit your way to AI fluency across a workforce of 2.23 million because there aren’t enough such people, commercial tech outbids you for the ones who exist, and the clearance pipeline slows everything else down. You can train your way there, if you are willing to stop treating the credential as the primary filter.
Then, there is a second reason this is urgent, and it makes the case for training rather than hiring even stronger.
Consider a technician who has spent thirty years learning the small signs that a component is starting to fail. Some of that lives in manuals and maintenance records. Much of it lives in judgement: what to look at, what to ask, which anomalies matter and which are noise.
AI cannot manufacture that experience, but it can help capture and distribute it. Technical documentation becomes searchable rather than archaeological. Historical failure data surfaces at the moment somebody needs it rather than three weeks later. Training simulations reflect what actually went wrong on this airframe rather than a generic curriculum. The retiring technician’s knowledge no longer has to remain a single point of failure, and that changes what apprenticeship can be. A young technician no longer has to spend two years learning how to find information before learning how to apply it. Better tools can compress the information-retrieval part of the learning curve, leaving experienced people more time to teach judgement, which is the capability technology cannot simply download.
None of that happens without deliberate investment, and it is worth speaking clearly about the failure mode, because aerospace and defence cannot afford a workforce that accepts machine-generated output uncritically. Human judgement becomes more valuable, not less, as the machines get more capable, which is exactly why the industry cannot afford to let its most experienced judgement retire uncaptured.
When I think about the remarkable story of the “non-lost” asteroids, the part I keep coming back to is that the objects were always in the picture, and what was missing was an instrument capable of seeing them.
In December 2023, the US Government Accountability Office published a review of how the Department of Defense manages its AI workforce. The finding was blunt: despite investing billions of dollars in artificial intelligence, the department could not fully identify who was part of its AI workforce or which positions required personnel with AI skills, and therefore could not effectively assess the workforce it had or forecast the one it would need.
I would encourage anyone in this industry to read that as a mirror rather than a headline. You cannot manage a skills transition you have not mapped. Most companies do not know which of their roles are most exposed to changing technology, which employees already hold transferable capability, or where institutional knowledge sits with three people who are all eligible to retire.
In April 2026, the Department of the Air Force approved an AI Hiring and Talent Development Plan built on recruiting, retaining and training. Its training standard shifts from course completion to proven skills and competencies for personnel in AI-related roles or seeking them. It establishes a baseline expectation of AI literacy across the force rather than confining it to a specialist cadre, and creates a Dual-Track Career Model allowing technical experts, including those in the Guard and Reserve, to advance without being pushed into management.
Notice what that implies — the largest and most procedurally constrained employer in the sector is moving towards a model that puts greater emphasis on demonstrated capability alongside traditional credentials. The emphasis on retention and training also makes the underlying point clear: existing personnel are a significant part of the talent pool it needs to develop.
Commercial aviation is pointing the same way, and Boeing’s own workforce outlook singles out competency-based training and assessment as central to closing the technician gap. Private industry has been slower to act on the same conclusion and has fewer excuses. A contractor can rewrite a job requisition in an afternoon. What usually happens instead is that the requisition describes the person who is retiring, which guarantees a search for a profile the labour market stopped producing fifteen years ago.
I want to be fair about the genuine constraints, because dismissing them is how outsiders lose the room. Clearance is real, and no training design fixes the queue. Export control narrows the pool before anyone assesses a skill. Safety-critical certification exists for excellent reasons, and nobody sensible wants a stress analyst with a twelve-week bootcamp behind them signing off on a wing spar.
None of that explains the degree requirement on roles where the degree is decorative. It does not explain why an avionics technician with eight years on the floor and demonstrated fluency in diagnostic tooling should necessarily have to go back to school before moving laterally into a data-adjacent role, or the persistent belief that a four-year degree completed in 2011 is stronger evidence of current capability than a validated skills assessment completed last month. The industry has world-class systems for verifying that a part meets spec and almost nothing comparable for verifying that a person does.
I work in healthcare and public-sector training, not aerospace, and I am not going to pretend those are interchangeable. What transfers is the mechanics of a shortage, which are consistent wherever you find one.
Training works when it is anchored to a specific employer’s specific demand, when the cycle is short enough that a working adult can finish it without leaving their job, and when the outcome is a verified skill somebody has already agreed to hire for. It fails when it is generic, long, and disconnected from a named vacancy. Earlier this year the State of Delaware put $2.3 million into free training for residents on our platform, which happened because the training was tied to jobs that existed rather than to a curriculum somebody liked the look of.
The aerospace version is not exotic. A mid-career technician completes a validated programme in the digital tooling their own employer already runs, is assessed on demonstrated competence, and has a defined role and a technical advancement track waiting at the end. That is not innovation, and the Air Force’s April plan puts many of those principles into practice.
I am 23, and I have watched this industry spend years describing people my age as the answer to its workforce problem while making itself progressively harder for us to enter. Both things can be true at once, which is why the shortage persists. Skills-based hiring requires an organisation to trust its own assessment more than a university’s, and large institutions are extremely reluctant to take on that liability.
Every year this goes unfixed, the backlog grows, the average age climbs, and more institutional knowledge leaves with nobody positioned to catch it. The people who could catch it are already inside these companies, or one adjacent industry over, doing work far closer to the requirement than any current requisition recognises.
The Asteroid Institute did not find those 27,500 objects because the sky changed, but because somebody built an instrument capable of recognising what was already in the archive. Aerospace has the archive. It has millions of people and a labour market full of adjacent talent it has trained itself not to see.
