Walk into any operations control centre in the world and you will see something remarkable: dozens of experienced people, surrounded by screens, making hundreds of small decisions every hour to hold a complex network together. Aircraft, crew, passengers, maintenance slots, airport constraints, weather — all of it must line up, in real time, every day. The people who do this work are some of the most skilled decision-makers in any industry I have come across.
So, when we talk about artificial intelligence in aviation, I think we sometimes start the conversation in the wrong place. We talk about algorithms, predictive models, dashboards, automation, and vendors. What we talk about far less is what decides whether any of it works: the organisation around it.
After spending much of my career building data, analytics, and decision-support capabilities for airline operations, I have come to a simple conclusion. The constraint is no longer access to technology. The technology is here, and it is getting cheaper and better every quarter. The real challenge is integrating intelligence into the everyday decision-making culture of an airline. That is a far harder problem, and it is the one that quietly decides which AI initiatives scale and which ones quietly disappear after the pilot phase.
The industry has always been data-rich — that isn’t the issue
Airlines have been data businesses for decades. We were early adopters of digital systems, we have enormous volumes of operational data, and our analytics maturity is genuinely impressive compared to many other sectors. A continuous-improvement mindset is part of the culture; we are always looking for ways to reduce waste, remove error, and tighten safety margins.
The arrival of modern AI — cheaper compute, better models, agentic tools — has rightly excited strategists across the industry. The opportunity to further optimise workflows, reduce risk, and handle disruption more gracefully is real. But excitement about technology has a habit of distracting us from the less glamorous work that determines outcomes.
A technically perfect prediction is useless if the organisation cannot act on it
This is where most aviation AI initiatives stall. Not because the model is wrong, but because the organisation around the model cannot move fast enough to use what it produces.
Different departments inside an airline are optimising for different things. Operations control is focused on schedule stability, passenger protection, and on-time performance. Maintenance is focused on keeping aircraft airworthy and protecting future availability. Crew planning is focused on legality and fatigue. Commercial is focused on yield. Network planning is focused on growth and connectivity. None of these objectives are wrong, but they are not always aligned, and sometimes they directly compete.
Layer on top of that the reality of how data is owned. In most airlines, each function controls its own data, defines its own KPIs, and uses its own language. “Delay,” “availability,” and “turnaround” can mean three different things depending on who you ask. When you try to build an enterprise-level AI capability on top of that fragmented landscape, the model is technically accurate but organisationally stranded — there is no shared definition of success, no single owner of enterprise optimisation, and no clear path from insight to action.
This is why I say the integration problem is bigger than the modelling problem.
AI adoption is a human-trust problem before it is a technical one
The other reality we must be honest about is human. Operations teams trust experience, instinct, and operational memory. They have seen black-box recommendations come and go. A controller deciding whether to swap a tail, hold a connection, or re-sequence a maintenance task is making a high-consequence call, often under time pressure. If the system cannot explain why it is recommending something, the recommendation will be ignored — and rightly so.
This is especially true in disruption management, tail swaps, recovery decisions, and maintenance prioritisation. In these moments, explainability matters more than algorithmic sophistication. A simpler model reasoning a controller can follow in ten seconds will beat a more accurate model whose reasoning takes ten minutes to unpack. Every time.
Digitising an inefficient workflow does not make it efficient
Another mistake I have seen often is digitising a broken process instead of redesigning it. A team builds a predictive alert, but the approval chain behind it is still manual, escalations still move slowly, and the action still depends on three people being available on a chat group. The insight exists earlier; the action does not happen any faster. We have spent money, we have produced a dashboard, and the operational outcome is unchanged.
This is why workflow redesign — not just tooling — must be part of any serious AI programme. If the decision cadence of the organisation does not change, the intelligence layer is decoration.
What human-centred AI looks like in practice
The most useful framing I have found is to think of AI in aviation as decision augmentation, not decision replacement. Controllers, dispatchers, engineers, and planners remain central. The best systems improve situational awareness, expand the range of options visible to a human, and shorten the time it takes to evaluate them. They do not try to take authority away from the people who have earned it.
In high-consequence operational environments, this is not a philosophical preference. It is a design requirement.
What the winning organisations will do
In my experience, the airlines that succeed with AI will share a few habits:
- They will align KPIs at the enterprise level, rather than letting each department optimise locally at the expense of the network.
- They will integrate operations teams and analytics teams into shared product ownership, instead of treating analytics as a service desk.
- They will build trust incrementally — starting with advisory systems, earning credibility, and only then moving toward guarded automation.
- They will redesign workflows, not just tools.
- And critically, they will have visible executive sponsorship. Without it, AI remains a pilot project.
The future is organisational intelligence
The next competitive advantage in aviation will not come from who has access to AI tools. Those tools are increasingly available to everyone. It will come from which organisations can integrate intelligence, trust, and operational execution into a single, coherent decision-making ecosystem.
That is a leadership problem, a culture problem, and a design problem — long before it is a technology problem. And it is the problem worth solving.
Walk into any operations control centre in the world and you will see something remarkable: dozens of experienced people, surrounded by screens, making hundreds of small decisions every hour to hold a complex network together. Aircraft, crew, passengers, maintenance slots, airport constraints, weather — all of it must line up, in real time, every day. The people who do this work are some of the most skilled decision-makers in any industry I have come across.
So, when we talk about artificial intelligence in aviation, I think we sometimes start the conversation in the wrong place. We talk about algorithms, predictive models, dashboards, automation, and vendors. What we talk about far less is what decides whether any of it works: the organisation around it.
After spending much of my career building data, analytics, and decision-support capabilities for airline operations, I have come to a simple conclusion. The constraint is no longer access to technology. The technology is here, and it is getting cheaper and better every quarter. The real challenge is integrating intelligence into the everyday decision-making culture of an airline. That is a far harder problem, and it is the one that quietly decides which AI initiatives scale and which ones quietly disappear after the pilot phase.
The industry has always been data-rich — that isn’t the issue
Airlines have been data businesses for decades. We were early adopters of digital systems, we have enormous volumes of operational data, and our analytics maturity is genuinely impressive compared to many other sectors. A continuous-improvement mindset is part of the culture; we are always looking for ways to reduce waste, remove error, and tighten safety margins.
The arrival of modern AI — cheaper compute, better models, agentic tools — has rightly excited strategists across the industry. The opportunity to further optimise workflows, reduce risk, and handle disruption more gracefully is real. But excitement about technology has a habit of distracting us from the less glamorous work that determines outcomes.
A technically perfect prediction is useless if the organisation cannot act on it
This is where most aviation AI initiatives stall. Not because the model is wrong, but because the organisation around the model cannot move fast enough to use what it produces.
Different departments inside an airline are optimising for different things. Operations control is focused on schedule stability, passenger protection, and on-time performance. Maintenance is focused on keeping aircraft airworthy and protecting future availability. Crew planning is focused on legality and fatigue. Commercial is focused on yield. Network planning is focused on growth and connectivity. None of these objectives are wrong, but they are not always aligned, and sometimes they directly compete.
Layer on top of that the reality of how data is owned. In most airlines, each function controls its own data, defines its own KPIs, and uses its own language. “Delay,” “availability,” and “turnaround” can mean three different things depending on who you ask. When you try to build an enterprise-level AI capability on top of that fragmented landscape, the model is technically accurate but organisationally stranded — there is no shared definition of success, no single owner of enterprise optimisation, and no clear path from insight to action.
This is why I say the integration problem is bigger than the modelling problem.
AI adoption is a human-trust problem before it is a technical one
The other reality we must be honest about is human. Operations teams trust experience, instinct, and operational memory. They have seen black-box recommendations come and go. A controller deciding whether to swap a tail, hold a connection, or re-sequence a maintenance task is making a high-consequence call, often under time pressure. If the system cannot explain why it is recommending something, the recommendation will be ignored — and rightly so.
This is especially true in disruption management, tail swaps, recovery decisions, and maintenance prioritisation. In these moments, explainability matters more than algorithmic sophistication. A simpler model reasoning a controller can follow in ten seconds will beat a more accurate model whose reasoning takes ten minutes to unpack. Every time.
Digitising an inefficient workflow does not make it efficient
Another mistake I have seen often is digitising a broken process instead of redesigning it. A team builds a predictive alert, but the approval chain behind it is still manual, escalations still move slowly, and the action still depends on three people being available on a chat group. The insight exists earlier; the action does not happen any faster. We have spent money, we have produced a dashboard, and the operational outcome is unchanged.
This is why workflow redesign — not just tooling — must be part of any serious AI programme. If the decision cadence of the organisation does not change, the intelligence layer is decoration.
What human-centred AI looks like in practice
The most useful framing I have found is to think of AI in aviation as decision augmentation, not decision replacement. Controllers, dispatchers, engineers, and planners remain central. The best systems improve situational awareness, expand the range of options visible to a human, and shorten the time it takes to evaluate them. They do not try to take authority away from the people who have earned it.
In high-consequence operational environments, this is not a philosophical preference. It is a design requirement.
What the winning organisations will do
In my experience, the airlines that succeed with AI will share a few habits:
- They will align KPIs at the enterprise level, rather than letting each department optimise locally at the expense of the network.
- They will integrate operations teams and analytics teams into shared product ownership, instead of treating analytics as a service desk.
- They will build trust incrementally — starting with advisory systems, earning credibility, and only then moving toward guarded automation.
- They will redesign workflows, not just tools.
- And critically, they will have visible executive sponsorship. Without it, AI remains a pilot project.
The future is organisational intelligence
The next competitive advantage in aviation will not come from who has access to AI tools. Those tools are increasingly available to everyone. It will come from which organisations can integrate intelligence, trust, and operational execution into a single, coherent decision-making ecosystem.
That is a leadership problem, a culture problem, and a design problem — long before it is a technology problem. And it is the problem worth solving.
