My role and contribution
At the start, we divided the project into parts. I took on the problem and opportunity section and the governance section. I began by researching and writing down everything relevant to the project in a separate document. This research was extensive. The document went through several iterations and retouches. For example, my first version of the problem and opportunity described the gap in quite general terms: Reddit has a lot of data, but brands cannot use it well. Later versions named Reddit’s existing product and the specific workflow we wanted to change, from keyword search to proactive monitoring. For governance, I worked through the privacy and legal side, using the course framework of the GDPR, AI Act, DSA and DMA, and the failure risks, such as false positives and the need for human review.
Because of this approach, my understanding of my own topics became very deep. The downside was that the final output, the slides, came at a late stage. After my own research, I also read through my teammates’ sections. I did some extra research to understand their parts better and gave them feedback. This helped me see how my sections connected to the rest of the project.
For the final deck, I was largely responsible for the layout of the slides. I was surprisingly fast at this, which is probably why I ended up doing most of it. I could turn a lot of research into something visual and scannable quite quickly.
Important decisions and iterations
Our project had two big pivots. The first was the choice of track. We started with Track B, a new venture. During the research this became confusing. The evidence I found did not seem to fit a new venture. It kept pointing to Reddit itself, since Reddit already owns the data and the community history. I raised this in a group discussion, and it turned out all of us had run into the same problem. Together we decided to switch to Track A and transform an existing organization: Reddit.
The second pivot I raised as well. While doing the slides, I found evidence that Reddit already has a Community Intelligence dashboard. We realised this on the day of the deadline. Our idea was suddenly not as new as we thought. We had to come up with a differentiating point quickly. Our answer was to move from “pull” to “push”. In Reddit’s current tool, the user has to know what to search for. In our version, the company pastes its website link and the platform proactively pushes the signals that matter to that company.
What surprised me or failed
What surprised me most was how quickly we solved both pivots once we were together. One of the better things we did as a team was meeting physically on campus. When we discussed a problem face to face, we came to solutions much faster than over chat.
Not everything went well. After the switch to Track A, my governance section still read like a plan for a new start-up. It described risks for a company that did not exist yet. I had to rewrite it so the risks belonged to Reddit, for example Reddit’s dependency on user trust and its legal disputes with firms that scraped its data. Some of my early research was therefore less useful than I hoped. What could also have been done better is starting the slides earlier. Being deep in research made us, or at least me, miss details early on. If we had built the slides alongside the research, we might have found the existing Reddit dashboard sooner.
How feedback and evidence changed the work
Evidence changed our work more than once. The research showed us Track B did not fit. The discovery of the Community Intelligence dashboard forced us to sharpen our idea. Feedback worked in two directions. My teammates’ comments on my sections helped me cut them down to what mattered for the slides, since my documents had much more detail than a slide can hold. When we checked our draft against the grading rubric, we also saw gaps. Our main claim, that semantic AI finds signals earlier than keywords, was not yet proven, and the prototype only used synthetic data. Instead of hiding this, we stated it openly in the governance and evaluation slides. This taught me that being honest about limits makes a strategy more convincing, not less.
What I learned from collaboration
Dividing the work in parts made each of us an expert on one topic. But the real progress happened when we came together. Brainstorming as a group turned problems into solutions quickly. Reading my teammates’ work also taught me that giving feedback works better when you understand their topic, not only your own. Next time I would combine the deep individual research with earlier group check-ins, so the whole picture comes together sooner.
How my understanding of AI and business strategy changed
What I learned is that developments in AI happen so quickly that you always need to keep yourself updated with new evidence. Something that looks new today may already exist tomorrow. It is also hard to come up with solutions that are truly new and not already implemented, especially by large companies. They probably have dedicated budgets and people to constantly follow new technological developments.
This connects to an idea from the first lecture: AI is an amplifier, not usually the moat. The models we used can be rented by anyone. What makes the idea work for Reddit is its scarce complement: its own data, history and communities. That is also why our final slides say the moat is Reddit’s context, not the model.
However, with good brainstorming and teamwork, it is also not hard to create add-ons that make an AI technology more attractive. You can tailor it to specific businesses and support their strategy. Our push model is an example of that. We did not invent a new technology. We changed how an existing one delivers value. Working on governance also showed me that a good strategy needs human oversight. In our design the AI proposes and an analyst decides, because a false alarm costs the customer time and trust.
The demo day was insightful. It showed how other teams approached their strategy. What I noticed most was that our team focused on a business strategy and on solving a problem for one company. Some other teams focused on solving a societal problem with the help of AI. Seeing this made me think about our own project differently. The benefit to society in our idea is indirect, for example when brands notice product problems earlier. Next time I would think about the wider impact from the start, not only about the paying customer. The choice of problem shapes the whole strategy.
