AI & Organisation Design
Enterprise AI can help airlines design organisations that serve customers better
Aviation is full of tricky acronyms. Do you know your FOHE from your PSC*? Or ITCM versus TSR**? Amsterdam-based KLM has probably the hardest of all to remember. Relatively easy “Royal Dutch Airlines” is a tougher “Koninklijke Luchtvaart Maatschappij” in the original Dutch.
* FOHE = Fuel Oil Heat Exchange, PSC = Passenger Supply Channel
** ITCM = Initial Technical Compliance Meeting, TSR = Test Seat Review
KLM has done well in the pandemic. In 2020 their revenue fell 54% but they were able to cut costs by 46%, limiting EBITDA to (€75m). A fully executable turnaround plan was developed in only ten weeks.
It was KLM’s ability to pivot it’s organisation and work with it’s stakeholders that led to these good results. Their organisation structure allows department heads to rationalise costs and maintains a corps of experts to accelerate change on the cost side. This is achieved by a strict hierarchy that keeps certain managers directly accountable for specific business issues.
The organisation is supported by three pillars. Sensible financial policies like paying almost its bills in 85 local currencies help to save millions. Strong network initiatives help too, such as using the airline’s relationship with Philips, a medical equipment and consumer electronics company, to create an air bridge with China. The final pillar is cargo and engineering – since the airline’s technical division service aircraft for third parties, this part of the organisation generates revenue even if KLM does not even operate a single flight.
The three pillars together supported a skeleton network for KLM during the COVID pandemic. They maintained service to 90% of European destinations, although with only 50% frequency. Sufficient density supported network reach. And almost no permanent staff were furloughed.
The Dutch flag carrier’s organisation comes with drawbacks though. Like Lufthansa, Cathay Pacific, Qantas, Air Canada, easyJet, Ryanair and other similarly organised carriers, they have isolated “factories” in the airline. And while these factories are great for cost control, they do not facilitate the commercial agility and innovation necessary for successful revenue generation. This matters because controlling costs can only impact profitability to the extent that the airlines – already lean organisations – currently spend money.
The upside of generating profitability through revenue growth however is effectively unlimited. And since almost all the factories touch or impact the end-customer they each have a role to play in this side of the profit and loss accounts.
Airlines have been making progress
After the 2007-09 global financial crisis the world’s most successful airlines were those that addressed their governance. Ed Bastian at Delta, IAG’s Willie Walsh (now head of the airline trade association IATA), Alan Joyce of Qantas and my personal hero Air Canada’s Calvin Rovinuscu made change happen by insisting on faster decision cycles and higher quality teams.
Madrid-based IAG, who own the Aer Lingus, British Airways, Iberia and Vueling brands among others, are particularly well-known for tight governance. Australian Qantas is too. They are tough negotiators and demand near-certain proof that anything they review will be directly actionable and quickly profitable.
But even within these airlines the various departments are inconsistent when it comes to enhancing profitability or customer experience. On FlyerTalk, an Internet bulletin board, it is a meme that any British Airways “enhancement” announcement will certainly cause pain for passengers.
Artificial Intelligence (AI) offers a solution, but not within the current structures
AI and machine learning are often touted as definitive solutions to airline challenges. Unfortunately based on my experience these technologies will fall short because in the current organisational environments even if decisions are data-driven the technology will be too laser-focused to spot the hidden patterns and insights that traditional computational methods miss because the goals are not holistically defined.
Current airline organisations will not meet future business needs or facilitate recovery post-COVID. At least, not if airlines are to reach their potential.
Airlines need better organisational analytics
At KLM and most other airlines the organisation is not based on data but what has been in place for years before, often many years. The organisation’s overall performance is not measured either – the accounts show profit or loss but are silent on how this was created by the company’s organisation.
Airlines will need to use organisational analytics to actively understand how their organisation impacts their performance if they wish to implement AI at enterprise levels. Fortunately AI can be used to develop the organisations themselves and I am convinced that this can be one of the first successful use cases for the technology in this industry (and other industries – even software companies can probably learn lessons too!)
What is the problem?
An organigram is not an organisation and it is not how people work either – real organisations are messy and people tend to talk to each other independent of reporting lines. But org charts do show how goals are transferred for execution – that is how the box-like factories become a problem.
The most important issue is that breaking down a complex process into existing departments leads inevitably to information loss. This leads to a further problem – when shielded from other units the framework of reference changes, meaning there is no logical sequential system and neither AI or machine learning methods can find the links that humans miss if another silo’s data is invisible to the algorithm.
We all know practical examples
Many of the challenges are in commercial. Traditional revenue management (RM) is about seats but other departments are about real customers. This is hard for RM practitioners to see unless they step outside RM and work elsewhere. Often, RM departments are defined by the current systems (and logic) they use.
For loyalty the co-branded cards are important financially, but since loyalty was traditionally about people who flew yesterday (rather than those who will fly tomorrow) many managers still focus on the programme members. The real customers of loyalty programmes are the banks who buy mileage currency to give to their customers.
As a result some emerging loyalty initiatives are a bit dodgy. Selling tier status (as opposed to earning it) stops people actually flying the airline and will hurt revenue in the long-term, since status-chasers are known to be less price-sensitive.
The economics of selling lounge access is also questionable – on short flights especially, priority channels and lounge access are a substantial part of the benefits that people with business class tickets pay for. Furthermore, overcrowded lounges degrade the experience for genuine regular ticket buyers and risks them flying with someone else. I often decide to dine at a decent restaurant rather than going into the lounge. Some airlines can send you an automated alert when the lounge is too full.
These trends are likely to get worse post-pandemic as airlines panic in their search for short-term revenue and push hard for any ancillaries they can. I often joke there are more people lining up at priority boarding than regular passengers.
Finally, each team has their departmental goals, and they are not built around any specific customer. There is no end model or even a common definition of who a customer is. Even at C-level conflicting objectives pull the organisation in all directions.
An analytics-driven prescription for commercial de-siloisation
Recent developments in business analytics have provided better insights into airline business units, enabling smarter and faster decisions with automation and removing repetitive work. Alaska Airlines uses machine learning to dispatch flights more efficiently, Air France uses AI to spot chances to save fuel and Southwest uses advanced analytics to complete network and schedule planning in real time with AI.
Commercial however remains relatively untouched by these developments. Their success depends on serving the needs of customers, not the mechanics of moving aircraft and should work backwards from customer experience rather than forward from an operating base. For this to work, airlines need consistent product and service delivery no matter whether the matter in hand is powered by sales, marketing or revenue management and this can only be done if the walls between these three silos are broken down.
I believe that advanced analytics offer the necessary insights that can lead to this de-siloisation in practice. Imagine the cost savings from removing silos and the value of new micro revenue models discovered by taking a more holistic view, driven by deep analytics? And going-to-market instantly through automation.
Making it happen requires multiple layers of deep learning to be combined on an AI platform, such as Databricks, and I am currently working on this enterprise flow for an airline project where I confirmed that when the goal is well-defined, automation is possible.
Using enterprise AI to design new airline organisations
So, how do we design an airline organisation that can best execute its strategy AND take advantage of Enterprise AI? By combining departmental processes and optimizing them around a common goal, to run faster and smarter. In other words, the solution lies in using Enterprise AI to guide organisation design.
Enterprise AI stacks departmental AI “problem solvers” (such as market analytics, trends, demand generation, ancillary, revenue management, fleet, finance, aircraft acquisition and slot requests) on top of each other and compares this to a higher overall logic (the “end model”) that integrates all workflows along a common goal. It takes the customer’s perspective of how airline products and services are received, perceived and used, as well as where and when. Then there is room for further monetisation of augmented data hubs.
The types of organisations that AI can design will simply be better for the customer, for example by enabling flights scheduling, ticket and other sales based on the day and time a passenger arrives rather than departs.
I foresee an airline organisation which changes the day-to-day tasks for most people working at airlines and where there are multi-disciplinary experts. That sounds more 2030 to me. Especially with Generation Z in the workforce.
There is plenty of room for commercial evolution and differentiation.
Do you want to find out more? Get in touch.
ricardo DOT pilon AT millavia DOT com