Introduction
As part of the AI Strategy Lab in our Information Strategy course, our team was challenged
to design a project using an emerging technology at the core of a corporate strategy (either
by improving an existing company through new solutions, or by creating a new structure
from scratch).
The goal of this project was to go beyond a purely theoretical vision of the course to
understand the real challenges related to implementing AI into a workflow, while
navigating strategic decision-making and the dynamics of multicultural collaborative
work. Multicultural, insofar as our group was composed of a Frenchman, a Pole, a Slovak,
and a Dutchman. This significantly enriched our project and our collaboration.
Summary of the final project
For our project, we worked on the management of flight cancellations at KLM at the
Schiphol hub, a problem that is costly to airlines in terms of EU261 compensation,
accommodation, working time, and loss of loyalty. Current solutions are either time
consuming (manual processing with queues) or unsuited to complex cases (algorithms
with inadequate rules).
Our solution proposes an Artificial Intelligence agent designed to handle the complex
cases that current systems fail to resolve. This AI agent reads the messages left by
passengers, identifies their specific constraints (requests to stay in a group, deadlines,
connections to respect), and integrates them into the existing booking engine to propose
better flight options.
Decisions, Obstacles, and Feedback
Regarding the choice of our project idea, we were quite fast in making our decision. We
had organized ourselves so that everyone thought of one or two ideas, which we
presented to each other so that the group could collectively choose the most relevant
one. We all had interesting ideas often linked to our backgrounds, whether personal,
professional, or academic.
For my part, I had submitted the idea of an AI-enriched database, to solve a problem I had
encountered a few weeks earlier during my internship at Rothschild and Co. Ultimately,
we unanimously turned to the project mentioned above because it was the most concrete
project, suited to the implementation of new technologies, and accessible to the greatest
number of people. My project, for instance, was too finance-oriented and therefore
abstract for my teammates unfamiliar with this field.
However, it was after a few days of reflection and research that we realized the presence
of numerous existing competitors and solutions already in place. This created hesitation
within the group, as we wanted to be innovative. We therefore took a step back in order to
reflect on the future of our project and studied the diAerent possibilities available to us.
After having determined two possibilities, we sought your help in order to ensure we
stayed within the initial alignment of the project and to be certain of its relevance. Your
extremely rich feedback allowed us to make a swift decision right afterwards in order to
continue our journey. We finally opted for a solution aimed at improving an existing
product, insofar as a solution was already in place within the airline. We can clearly see
here the importance of feedback : on one hand, the feedback between us, which allowed
us to identify blocking elements and try to resolve them with solutions, and on the other
hand, your feedback, which allowed us to decide by adding supplementary elements of
answers.
An extract of my contribution
Thanks to my previous background in finance acquired at NEOMA BS and the current
courses taken at Rotterdam School of Management, I participated in building the financial
strategy and the business model of our AI agent for KLM. We defined a hybrid model : for
internal use, the solution is not a classic SaaS $but an investment justified by the savings
achieved, with a goal of amortization over one to two years. For a potential distribution to
SkyTeam alliance partners, we did not choose to adopt the per-seat model, which is not
suitable since AI precisely reduces the need for staA as the model’s feedback loop
improves its performance. We opted for a pay-per-use pricing model facturated per
rebooked passenger.
We modeled the savings achieved by quantifying the reduction in working time needs,
care costs, and EU261 compensations. To ensure the accuracy of the calculation, we
integrated the variable costs of the artificial intelligence API tokens into the equation,
while comparing them against the residual cost of escalations to human agents. The
weakness of our modeling lies in the fact that perhaps tomorrow the cost of tokens will
be multiplied by 10, this possibility poses a viability risk to our project.
I also focused my involvement on the Front-end part and the eAiciency strategy. Initially, I
preferred that our team focus on building the user interface, rather than on the
development of a fully functional AI. Indeed, the skills required for the front-end and the
integration of an LLM API are quite diAerent, and a perfectly polished AI was not formally
required.
However, after discussion, we agreed to develop both, as designing the AI allowed us to
experiment and build our technical skills. This option of doing everything holds up if we
have surplus time. On the other hand, if our available time shrinks, we will have to switch
to a “Lean” approach, where we will not use any time for what is not strictly essential, in
order to guarantee a viable deliverable for the presentation.
This part of the project was interesting, as it allowed me to mobilize cross-functional skills
acquired throughout my background to meet a challenge combining strategy, technology,
and finance.
Our collaboration and its learnings
Within our team, the diversity of our profiles was an asset. My skills in simple
development, notably for the front-end, and my background in finance helped structure
the financial side. These skills articulated perfectly with those of my partners. Their
expertise in marketing guaranteed a user-centric approach, very important for the
passenger experience, while their skills in Business Information Management enabled
rigorous and structured management of data and information flows. This mix of a
technical/financial approach with a user-oriented vision and data organization created a
true synergy, making our solution both strong and relevant.
Conclusion
This project was a good opportunity to give meaning to the concepts covered in class. The
theory on corporate strategy was useful when we had to adapt our initial idea in the face
of a competitive market. So we understood the importance of feedback and proper
positioning.
However, the biggest lesson I take away from it is about AI implementation. Before the
project, I saw AI as a feature that you add to a product. I understood that its integration
completely modifies the workflow. Setting up a solution requires managing the
complexity of data flows, hidden costs, and ensuring that the technology answers a real
strategic need and not just a simple passing tren
