AI in UK Passenger Transport: Rail, Aviation and Maritime

Why this briefing

Disruption is rising and compensation has become a volume business. Rail operators closed 3.8 million delay compensation claims between April and July 2026, 55 per cent more than a year earlier, as cancelled trains rose by a third. In aviation, 51,691 disputes reached the approved dispute resolution bodies or the CAA in the year to March 2026, and in 45 per cent of those decided the passenger obtained a remedy.

The passenger with an assistance booking is the one the system most often fails when things go wrong. The rail regulator's own research found 86 per cent were met when their service ran normally and 67 per cent when their train was cancelled, and concluded that Passenger Assist "is more unreliable when services are disrupted". The aviation regulator reached the same view from different evidence.

AI is arriving at exactly that moment. The rail industry's AI plan sets out to automate compensation eligibility and already uses machine learning to screen Delay Repay for fraud; a search of its text finds no mention of disabled passengers. The EU is automating air compensation claims by law from October 2027. The rules already draw the line: rail redress for a failed assist must be decided case by case, and since February 2026 automated decisions based partly on health data, such as an assistance booking, are tightly limited.

Much of the AI already in use sits in the operation the passenger never sees. The rail industry plans one incident prediction shared across control, customer information, crew and rolling stock; predictive models for doors, HVAC and traction aim to cut avoidable cancellations; AI cameras detect trespass and passengers in distress; Heathrow is using AI on stand cameras to speed turnaround. British Transport Police publishes every live facial recognition deployment it makes at stations, with its outcome. The industry's own plan warns that many AI pilots struggle to produce measurable financial benefit.

The human moment

The passenger whose assistance booking does not survive the cancellation. A wheelchair user has booked help at both ends of a rail journey. The train is cancelled and the passenger is rebooked onto another service, by a person or by a system. Whether anyone is waiting at the other end depends on whether the booking travelled with them. The regulator's own survey shows that when trains are cancelled, a third of passengers with booked assistance are not met.

What's inside

Seventeen pages, written for a transport operator executive who knows the passenger business well and AI less well:

• A note on the numbers — what is included, what was excluded and why, including an unattributed 93 per cent claims-automation figure
• More people travelling, and most of it measured
• The pressure point — the claim gets faster, the passenger who needs help does not
• Compensation as a volume business, first answers that do not stand, and the thinnest record, at sea
• Where AI actually is, and four risks specific to passenger transport
• Beyond the passenger — AI across the operation: recovery, planning, maintenance, security and turnaround
• The threat side — fraud against compensation schemes, and the model that answers it
• The UK legal position, including two bills before Parliament
• The change nobody has framed — the assistance booking is health data
• Beyond the UK — the EU automates the claim by law
• Seven things an operator can do now, none needing new law
• Three levels defined by where accountability sits, three horizons, and an eight-workstream way in

The discipline behind it

Every figure carries a source: the Office of Rail and Road, the Civil Aviation Authority, the Maritime and Coastguard Agency, the Department for Transport, official statistics, Parliament, a court, or an operator on its own record. Where the briefing reports an absence, such as words missing from the rail industry's AI plan or enforcement data not published, the source was searched for it.

The most quotable numbers in the sector are excluded on the record: an unattributed case study claiming 93 per cent automated claim processing and 70 per cent fraud-detection accuracy for UK train operators, and an airline's minutes-of-delay-saved figures given at a conference without a method.

Who it's for

A chief executive, customer director, chief information officer or accessibility lead at a train operator, airline, airport or ferry operator who has just picked up an AI programme, or a board member asked whether automated compensation and disruption tools can be shown to work for every passenger. It assumes the passenger business is well understood and AI is not.