Before the course, I assumed that GenAI was mostly about lowering barriers to entry, through reduced costs, less manual labor, and easier access to information. However, while working on a GenAI extension for IKEA Kreativ, I came to see that in the right ecosystem, GenAI can also raise entry barriers and help incumbents strengthen their competitive position.
Our project, RoomMate, extends IKEA Kreativ, the existing room-planning tool, with GenAI. Many customers struggle to compose a complete room in their preferred style. The many options to choose from often lead to decision paralysis, causing purchases to be postponed or abandoned. RoomMate addresses this by generating a styled room from a user prompt.
Within the team, I was responsible for the economic analysis and the competitive and ecosystem analysis, while also keeping the overall story coherent. At the start of the project, I drafted a rough storyline for our deck, leaving room for revisions and additional slides. This paid off when the deck was nearly complete, as I noticed that the narrative jumped straight from the problem to the business value without ever explaining the product itself. Adding one slide fixed that and made the story much easier to follow. It reminded me how important it is to take the reader’s perspective, especially when I was deep in the details.
One of the most important decisions that we made was about the revenue model. We considered two different options. First, charging customers per design, which would have given us a more steady, predictable income, but would also have created friction and slowed down adoption. The alternative was to make RoomMate free to use and earn revenue from the additional furniture and accessories customers would buy based on the recommendation of a fully styled room that matches their preferences and is ready within minutes. We chose the second option, assuming that capturing market share and keeping customers on our platform from first idea to final purchase would outweigh the income from per-design fees.
What I underestimated was what this choice meant for my part in the project. We traded a certain, easy-to-calculate income for one that depends heavily on assumptions: will people buy from an AI-generated tool, how many more will buy, and how much more will they spend? Every number had to be researched and put in the context of RoomMate. This taught me that the strategically attractive choice can also be the high-risk, high-reward one, which is harder to prove, but potentially far more valuable if it works.
The first version of my economics was far too confident. I had researched how many people use Kreativ and assumed that, with a GenAI add-on, many of them would start using RoomMate immediately. However, further research showed that add-on features have a much lower adoption rate, and that adoption builds up gradually. I therefore lowered my adoption assumption and assumed slower growth. This reduced revenue in the early months and moved the break-even point back. It also changed how I thought about adoption. Customers do not adopt a feature simply because it is available, new behavior takes time. Value for the customers, and therefore for IKEA, only appears when the tool is actually used. The pilot became far more than a financial check, as it is where we test whether people really use RoomMate and whether it creates value.
One finding I hadn’t anticipated was just how sensitive the revenue model is. Small changes in the share of users who buy, or the extra amount they spend, or IKEA’s profit per item sold moved the break-even point anywhere between 11 and 24 months. I also consulted my colleagues and went through a couple of iterations to make the numbers as realistic as possible.
However, “realistic” in my own eyes, did not mean the numbers were right. When I had finished the economics, a colleague went through all my figures. She tested whether I had set everything up correctly and whether someone who had not worked on it could follow the logic. She pointed out that she would calculate the minimum pilot result required for a “go” decision differently. She used a different assumption, which was better substantiated than mine. We therefore adopted her recalculation, which made this part of the analysis even stronger. This taught me the importance of working together, asking questions and having to explain your work exposes weaknesses. It helped me improve and learn, which led to a better final analysis. Next time, I would involve a colleague earlier to review my numbers, so wrong assumptions or errors are caught sooner.
At the start, I knew that AI has a major impact on companies: fewer people need to be hired, efficiency rises, costs fall, and new markets become easier to enter, which increases the threat of new entrants. What surprised me, however, is that GenAI can also have the opposite effect. The reason is that GenAI is not the moat. The moat arises because IKEA’s products, ecosystem, complementors, and assets together allow it to use GenAI in a way that builds a cumulative competitive advantage over time. Every time customers use the tool, IKEA gains data, which can then be used to improve the tool, creating more value for customers and attracting more of them. This is called the learning loop. On top of that, our product is an information good. For ordinary products, further growth eventually stops paying off, because of diminishing marginal returns. Information goods, however, work differently: once the tool is built, serving an additional user costs very little, which makes it highly scalable. Combined with the learning loop, this allows IKEA to build an advantage that new entrants and competitors will find hard to match. However, this advantage is not automatic, as the learning loop only starts if customers actually use RoomMate. That is the assumption my economics showed to be least certain, and it is what our pilot has to prove.
This also showed me how important it is to start from the customer’s problem. Using AI only because of the hype will not create a sustainable advantage, because competitors can do the same. We should ask ourselves: which assets and complementors do we already have that, combined with AI, can solve the customer’s problem better, cheaper, or faster? With that view, AI can help a company thrive in a competitive market. For RoomMate, that meant starting from the customer who gets stuck choosing, not from what the technology can do.
