Throughout the eight-week project, my main responsibility within Group 37 was coordinating and facilitating, combined with my ownership of the strategic defensibility of our proposition.From the very start of the project, I took the initiative to get our group together for onlinemeetups to discuss project milestone progress. With four team members balancing different project views and tight course deadlines, delegating tasks evenly among Juline, Mateo, Gia and myself was essential to ensure a strong outcome.
Personal Contribution
In terms of content, I concentrated my efforts on two critical areas: Ecosystem & Defensibility (slide 7) and the Future-Market Test (slide 11). My objective was to validate our solution using theoretical frameworks provided in the course. With that I examined specifically how an AI-enabled matching solution compares to incumbent listing platforms, how it defends against platform envelopment and how Stekkies can prepare itself for the emerging paradigm shift toward autonomous agentic commerce. The most critical learning milestone of our project journey occurred relatively early. In our initial proposal we designed an overly ambitious platform called ‘HomeScout’. It was projected to be a comprehensive search assistant for students that attempted to solve most of the current international student rental market friction (e.g., scraping listings, detecting fraud, forming rental groups). The initial proposal was rejected by the course instructors due to the excessive scope. In hindsight, I completely agree with the view of the course instructors; we had overcomplicated it for ourselves when we tried to come up with a tool that attempted to fix the entire rental journey. After the lecture, I got our group together for an online meetup where we eventually agreed that we needed to focus on a specific problem. By applying the concept of value chain decoupling, we came up with a solution that targets one of the biggest bottlenecks in rental markets for international students: forming credible multi-tenant sets to unlock a range of previously unreachable multi-bedroom properties. This narrowed our focus from an overcomplicated search assistant to an impactful innovation: Stekkies Together. Within our narrowed approach, I took ownership of designing our defensibility architecture and mapping of uncertainties. Rather than just using the framework by Dimoka et al. (2012) as a theoretical checklist, I used it to diagnose why matching alone would fail in the physical rental market.
• Product Uncertainty: Our first prototype test run provided valuable feedback/insights into product uncertainty. The algorithm matched 10 out of 94 groups into apartments where municipal permit regulations forbid multiple tenants. We resolved this by implementing a permit screening that verifies whether properties legally allow three or more non-family renters under municipal rules.
• Fit Uncertainty: We addressed this with our AI-driven grouping tool based on budgets and personal preferences.
• Peer/Seller Uncertainty: We minimized this through upfront identity verification and solvency checks, which serve as observable signals for the market that should reassure landlords and brokers.
I also analyzed the threat of platform envelopment by dominant portals like Pararius or Kamernet. I found out that existing portals face significant constraints: their business model partly relies on charging broker listing fees and their websites only seem to gather shallow search data. Attempting to bundle our group-matching approach would affect their broker relationships. Furthermore, by applying Yerramilli-Rao et al. (2025), I demonstrated that algorithmic matching alone offers no long-term defensibility. Our true competitive moat lies in scarce complements such as proprietary data from matches, resolved disputes, verified identities and established trust with brokers/partners. My second major analytical contribution was structuring the Future-Market Test using the agentic strategy frameworks from session 8. Applying Walsh’s (2026) agent-firm-market framework forced me to look at the future of Stekkies’ current model. I learned that Stekkies’ current business model will become completely obsolete once seekers deploy autonomous agents that monitor and filter listings. To prepare for this shift, we proposed that Stekkies transform into an ‘Agentic Rail’. In this model, Stekkies authenticates agent identity and renter solvency, ensuring broker bots only interact with pre-verified, credible tenant sets. Furthermore, I insisted on establishing a strict Human-in-the-Loop boundary to prevent automated agents’ commitments and liability disputes.
Collaboration and Personal learnings
Facilitating our team provided valuable lessons for future projects. Structured communication especially helped us to turn around the project from a large, unfocused scope to a cut-down, workable innovation. Overall, we’ve had a very productive period. All team members have delivered project milestones on time, and we’ve been able to have objective, unfiltered discussions about Information Strategy topics. A key debate arose around data collection for our matching tool: we were discussing possible data points that could be collected to optimize algorithmic accuracy. One of the team members proposed the idea of gaining a very deep understanding of renter identity and university background. We then got into a discussion about bias in the algorithm and the possibility of it duplicating the discriminatory patterns that currently exist in the Dutch rental market. Through structured discussions we came to a consensus that we must restrict the data inputs to non-sensitive lifestyle attributes, ensuring our project fit within responsible AI principles.
Thoughts on AI & Business Strategy
I started this course thinking digital innovation was mostly about advancements in technology. Building Stekkies Together proved me wrong. Our first idea, HomeScout, was technically impressive, but it tried to do far too much. It was the narrower Stekkies approach that made sense as a business. My takeaway is that a good digital strategy depends on how you design the market and who has an incentive to take part. The algorithm matters, but it is easy to copy. What is hard to copy is the trust and data we try to build around it.
