My role and contribution
The AI Strategy Lab was a true challenge and learning opportunity in terms of communication, teamwork, and collaboration. Working on a project involving unfamiliar technical concepts pushed me outside my comfort zone and encouraged me to identify how I could best contribute to the team.
Since I lack a technical background, I did not take on the responsibility of building the prototype. Instead, I took it upon myself to help coordinate our actions by dividing responsibilities, identifying and communicating the next steps, and ensuring that we maintained consistency across the different parts of our work. I was also responsible for developing the Governance and Failure Analysis and Final Recommendation sections, which required me to consider not only the potential benefits of our solution but also its limitations, risks of inaccurate demand forecasts, the need for human oversight, and conditions for implementation. Interestingly, my limited technical knowledge sometimes became an advantage. By asking my teammates to explain technical decisions and questioning whether certain functionalities were necessary or feasible, I encouraged discussions that helped us clarify our approach and direction. For example, by initiating a discussion about the respective roles of GenAI and agentic AI in our solution, we came to the conclusion that agentic AI was not actually necessary.
What surprised or failed
Initially, I was surprised by the criticism we received since we considered our idea both logical and realistic. However, as we worked through each point of the feedback, I began to understand that we had focused too much on what AI could do rather than on the specific problem it should solve. Looking back, I realise that I had accepted the assumption that greater automation would create more value without questioning it sufficiently. The feedback helped me recognise this limitation in my initial approach and reconsider my assumptions.
Another aspect that surprised me was that the information we wanted to use to improve demand forecasts could also make them less reliable. Our simulation showed that when online hype failed to translate into actual purchases, our solution risked recommending excessive stock, reinforcing the very problem we intended to solve. This made me realise that adding more data does not necessarily lead to better business decisions. What matters is whether that information is relevant, reliable, and interpreted correctly.
How feedback and evidence changed the work
The turning point in our project was the feedback we received on our initial proposal. Our idea was too broad, too automated, and unnecessarily complicated. We intended to implement AI functionalities that were already standard in inventory management systems, and we targeted an overly broad audience: organisations that held physical stock. In response to the feedback, we narrowed our target audience to one specific fashion retailer, LOAVIES, and focused on testing one assumption: whether adding external trend signals to historical sales data could improve demand forecasting and subsequently lead to better inventory decisions.
Our research also showed us that, although our idea had potential, other organisations were already offering similar services, such as trend forecasting and internal inventory planning. We therefore decided to differentiate our solution by combining both capabilities within LOAVIES and creating a retailer-specific learning loop.
Moreover, by testing our prototype against a simpler baseline alternative consisting of historical sales, we were able to assess whether adding an AI forecasting layer was justified. This also allowed us to identify the limitations of our solution and get a better understanding of where AI could actually create value.
Important decisions and iterations
One of the most important decisions we made was to reconsider the level of AI autonomy required in our solution. As the project progressed, we realised that not every part of the process required automation. A simpler, more controlled approach would allow us to focus on improving demand forecasts while leaving the final decision to merchandise planners. Working on the Governance and Failure Analysis section made it clear to me that human oversight was necessary since our system could face failures such as processing unreliable trend signals, providing inaccurate forecasts, or giving inappropriate recommendations, which would affect inventory decisions. This made me understand that the value of our system did not lie in replacing human judgement, but in making it more informed.
What I learned from collaboration
Before this project, my understanding of AI was largely theoretical. However, working with teammates who had more technical expertise helped me understand the system’s underlying logic, including how APIs, LLMs, and different software components could be used together. As I became more familiar with these concepts, I felt more comfortable questioning technical choices and discussing how they related to our business objectives. This was a real learning opportunity that sparked my curiosity and made me want to deepen my knowledge of how AI can be used to address real challenges.
Moreover, this experience highlighted the importance of combining different skills within a team. Collaboration is not simply about dividing tasks according to individual strengths, but also about making sure that the different contributions fit together into a coherent project. Because different teammates were responsible for the prototype, business analysis, economics, and strategic recommendations, we had to constantly ensure that our conclusions remained congruent across these components. At times, different interpretations of our solution led to inconsistencies between the technical design and our recommendations, which required us to revisit decisions together and clarify our shared understanding. This reinforced the importance of regularly communicating our progress and questioning whether our individual contributions still supported the same overall objective.
How the project changed my understanding of AI and business strategy
Throughout this project, I developed a better understanding of why simply using AI does not necessarily create a competitive advantage. Any business can use AI. The real advantage comes from how it is used and implemented, whether it actually improves a specific business process or decision, and the extent to which it can increase operational efficiency and financial performance.
I also developed a better understanding of the distinction between creating value and capturing value. Even if AI can create value, such as by improving forecasting accuracy, that value may not be captured if it does not translate into financial benefits that justify the cost of implementing and maintaining said system. This made me perceive AI through a different lens by approaching it less as a technological innovation and more as a tool for strategic improvement whose value must be demonstrated.
Therefore, if I were to approach a similar project again, I would first identify which business decision needs improvement, define how success would be measured, and focus on the simplest way to address it, rather than pursuing an ambitious automated solution from the get-go. I would work the other way around: start small and scale only if and when there is evidence.
