1. What did we build?
Stekkies Together is a Track A project to transform an existing organisation, by selecting an existing service and revolutionising it with agentic capabilities and delivering greater value to the user. We chose Stekkies, a Dutch housing-search service for young adults looking for rooms in the Netherlands. Despite its usefulness, we found it to be limited given that there are many more multi-room apartments than studios in the Randstad area. This means there often isn’t an easy way to find roommates, and so many students and young adults have to pass up on rare housing opportunities. Stekkies Together fixes this with a network to find roommates on the Stekkies platform, and maximise the chances of getting a good apartment.
2. What did I contribute?
From our initial pool of ideas, I contributed the basis of the one that eventually became Stekkies Together. I started this project with research into the biggest pain points for (international) students in the Netherlands, and found the housing crisis to be the most important and relevant idea. From here, I developed my pitch into a Track B combined agentic housing-search and roommate-matching venture, although it was seen as too broad in scope and overly ambitious in terms of feasibility. My teammates then suggested folding the idea into an existing housing-search platform, Stekkies. This solved both our issues, as it removed half of the project’s scope and also reduced the network effect problem given Stekkie’s existing user base.
After pivoting to Track A, we divided the tasks and I set out to develop the demo/prototype for the platform, along with its evaluation. This entailed agentic development and brainstorming to determine the platform architecture design and logic. I gave this first version to Gia, my teammate, who iterated on my ideas until we got to the finished product. My work from this point onwards consisted of adapting the demo to the presentation, ensuring we could easily and clearly communicate the functionality and benefits of our idea to our professors and coursemates.
Furthermore, I worked on the prototype evaluation, comparing a simulated rental market with and without Stekkies Together to show the efficiency benefits of finding rooms and roommates at the same time. From here, I began analysis of the prototype’s success criteria and whether we had met the necessary goals for the simulation and concept. Lastly, I worked on the failure analysis, our mitigations, and our learnings from them. This involved detailing our prototype iterations and the assumptions we employed.
3. What did I learn?
This project helped hone both my business and technical skills. On one hand, I learned to streamline, iterate and redesign ideas and systems, in order to make projects more feasible and realistic. This is a key skill to have; balancing time, effort, and scope is a difficult task but is very important when designing information systems and applications.
On the other hand, I also improved my technical skills as I worked primarily on developing the prototype for Stekkies Together. This included agentic programming with Codex, but also figuring out an appropriate way to evaluate the performance of our project in a simulated housing market as well.
Lastly, this style of group coordination was also new to me; in my Bachelor’s, it was common to do teamwork with friends and so it was easy to coordinate and collaborate, especially in person. In the Master’s, most students have jobs and other responsibilities meaning we had to learn to work with new people, and to also collaborate with more efficient and effective communication.
4. Important decisions, iterations and crossroads
Our pivot away from an overly ambitious and wide-scope project was the first big decision we took. It made the project significantly simpler, while still keeping the key benefit of roommate-matching improving the effectiveness of a housing search for young adults in the Netherlands.
Another crossroads was the process design; whether to first find a home, and then roommates or to find roommates and then a home. We first chose the former, thinking it would be the most efficient option for prospective tenants, but upon testing, we realised it made little sense. This would have drastically increased the amount of applicants per accommodation, as anyone could match to any home and just find roommates to fill it up later. Instead, we reversed the system, to first find matching roommates and then search for a slightly narrower subset of accommodations in order to maximise the total market efficiency.
5. Unexpected results
The first runs of our demo produced confusing results, as it gave almost identical results with and without Stekkies Together (that is, both market simulations made it difficult for tenants to find accommodation). This surprised me, as I thought our team would have to pivot the project again given the poor result of the demo. However, after inspecting the market conditions, I realised I had been too strict with the roommate-matching system, as people were too often incompatible due to minor inconveniences that didn’t reflect reality. After tweaking the conditions, Stekkies Together performed much closer to what we had forecast; a significant effectiveness boost for prospective tenants to find housing together. This iteration also helped to push our simulation to be more realistic by approximating the virtual users to rational agents, which is what humans tend towards (at least through bounded rationality).
6. Implementing feedback
As mentioned before, listening to feedback and implementing it was key for the development of our project. After coming up with the initial agentic house-search idea, I workshopped it with my teammates, who initially found it to be too similar to Stekkies’ business model. Nevertheless, we pushed forward, thinking that an ambitious proposal would be the best choice. This is when we got the professor’s feedback who confirmed that our scope was too wide and unfocused, which would have made the project much more difficult. From here, we reworked the idea into a Track A proposal under Stekkies, both using the feedback from our initial group brainstorming and from our proposal to the lecturer.
7. How did it change my understanding of AI and business strategy?
Lastly, this project helped change and reshape my understanding of AI and business strategy significantly. With social media and economic hype cycles, it’s easy to believe that combining anything with AI (whether it be an existing product or a new one) instantly makes it attractive, desirable or useful. What Information Systems taught me (through the lectures, readings, guest lecturers and the project) is that AI, like any other technology, has to be thoughtfully implemented and used to actually create more value for the user and capture more value for the seller. This became really clear in our brainstorming phase, when we went through dozens of product ideas with AI integration that simply didn’t provide clear value capture, delivery, or creation.
In simpler terms, it means that brainstorming and coming up with proposals for the project was difficult, as most of our ideas lacked in feasibility. Implementing AI into a robust business strategy was a lot harder than expected, but I learned a lot about how to combine the two into a feasible product, beneficial for both the seller and buyer. All in all, the project helped me gain more critical thinking skills to design and evaluate information systems in a more effective way.
