American Airlines seems to be behind in artificial intelligence. At Skift Aviation Forum on Wednesday, American’s Chief Operating Officer said that he doesn’t expect headcount reductions from artificial intelligence (like United does). American has more employees than any other airline in the world, so probably has the most room here. A function like finance surely has room for streamlining. AI is better than human beings at reconciling accounts.
His key example of how American Airlines uses AI today is the HEAT tool to management delays and cancellations during major events.
- As I understand it, this is really machine learning rather than classic AI as we currently think about it. It isn’t the use of large neural networks or classic supervised learning. As I understand it, it’s more of an optimization model glued to some heuristics… a predictive model and recommendation system.
- The airline has been promoting the tool for three and a half years.
The lack of plans for AI helping the airline reduce headcount, and their key artificial intelligence initiative being something that’s been in place for over three years, comes in the context of:
- CEO Robert Isom blasting Delta for using AI in ticket pricing.
- The airline increasingly outsourcing its IT function.
It makes sense that airlines are going to be slow to adopt AI. They’re among the most heavily regulated industries. Two pilots in the cockpit is required, even if AI can improve safety versus a human co-pilot. One per 50 passengers is a flight attendant rule. There’s intense lobbying over requirements which will prevent labor cost savings. Given the safety consequences of operational changes, airlines will be conservative anyway. They’re heavily unionized, capital-intensive industries.
But airlines are making some progress here. Delta, most famously, is increasingly turning over its pricing to AI. And United has shed staff at headquarters and will continue to do so. However the opportunities here are tremendous.
- Contact center / reservations / social care. A large share of chats, emails, and calls finally get handled by computer and customers won’t know the difference. Rebookings within fare rules can easily be handled this way, not just seat changes, baggage tracking, and issuing vouchers during irregular operations. Only edge cases will need escalation.
- Back office operations from accounts payable and receivable to revenue accounting, document extraction and simple actions will have machines replacing agents manually keying things in and doing first line review. Already “offers & orders”/modern retailing pushes automation in revenue processes.
- Marketing/CRM/loyalty servicing. An area closest to my own heart, drafting marketing copy, segmenting customer lists, and handling everything from FAQs to terms and conditions should be handled primarily by AI. Humans will focus on partnerships, while GenAI will measure campaign performance – even testing copy against human response before it gets sent.
- Revenue management analysts. Delta already says that this is happening so that prices update on a virtually continuous basis. Human revenue management, in their telling, has been a bottleneck. Computers do better at demand forecasting and pricing, with humans shifting into broader strategy and guardrails on the AI’s price-setting. Politics is a barrier here with concerns over individually-tailored prices though concerns here, at least in the medium-term, are largely overblown.
- Network planning & schedule optimization. Figuring out how to optimize aircraft utilization, and how to handle this against fleet maintenance, airport curfews, and network connectivity will be better done by computers.
- Crew pairing/rostering and disruption recovery. When Southwest melted down at Christmas 2022, their binding constraint was compute. So, too, with Delta trying to recover from the Crowdstrike meltdown. They couldn’t rebuild their schedules. Southwest operated at a portion of its schedul efor days as they handled re-pairing crews with flights manually. Combinatorial solvers and learning systems will optimize this in minutes, while internalizing union work rules. David Seymour says they’re better at operational recovery than anyone else (his example was CrowdStrike, saying they were just as affected as others, but that it isn’t really true).
- Maintenance planning & technical publications search. Shifting to predictive maintenance and tracking will mean fewer maintenance analysts. It won’t mean less maintenance, and probably not fewer maintenance techs – for awhile – given the long aircraft replacement cycle.
This doesn’t even get into airport operations and gate management, baggage and virtually any activity on the ramp where it’s ultimately robots that may replace people.
The truth is that technology allows us to really rethink the premise of air traffic control, but that’s a government and safety function so that, too, will tend towards inertia.
I single American out here because of David Seymour’s comments about American as though they’re a leader in this space. He got the question, and he flubbed it (and several people in the audience I spoke to agreed). But the airline industry as a whole needs to embrace AI more – not just American.
AI is better than customer service in India.