Project Learning Blog (Team 35): If anyone can build it, why build it?

When I started this assignment there was one question that kept going round in my head. So many people have access to AI and technology nowadays that it is not special anymore to build something in a short period of time, even if you don’t know a lot about the subject yourself. So how do you still come up with an idea that is really new and valuable? I started the assignment with that feeling and to be honest, it never completely went away.
The idea for ReUnite came from one of my teammates. ReUnite is an AI-assisted lost-and-found service for hotels, which matches the description a guest gives of a lost item with photos of items the hotel has found. Staff always check a match before anything is returned. When I heard the idea I recognised myself in it straight away, because I once left something valuable behind in a hotel myself and I could imagine how hard it is to get it back from a big chain. The idea felt useful to me, but I was not sure if it was special.

The problem
I wrote the problem and opportunity section and this was the part where I felt most sure about what I was doing. I made my own rough estimate of what lost and found costs a hotel, in a sector where there is already a lot of pressure on staff. With around 100 items a month and 10–15 minutes per item, a hotel spends about 17–25 staff hours a month on it, which comes down to €400–€625. For the Which? undercover test at seven UK hotels (Rhodes, 2026) I wrote that we treat it as “illustrative rather than representative”, because with such a small sample that is what it was.
Knowing how strong my numbers were gave me confidence, but at the same time it made me doubt. €400–€625 a month is real money, but it is also not that much, and a problem can be real and still be small. In my eyes the bigger opportunity was the value for guests. Fewer than half of guests (46%) who left something behind got it back (Rhodes, 2026), and that is where we could really make a difference.

Competitors
Then we found Loggit, a company that was already using AI to match lost items. That felt really bad, it felt like the whole idea was falling apart. This was exactly what I was afraid of at the start of the assignment, that the moment you start looking, someone is already doing it.
In the business model I wrote “Loggit already matches with AI, so advantage must come from data.” We positioned ReUnite as a white-label product that connects to the property management system (PMS) of a hotel and combines AI matching with integration and verification. Barney and Reeves (2024) argue that AI cannot give a lasting advantage on its own when everyone has access to it, and that the advantage comes from what you combine it with, for example data or processes. That argument really hit home for me, because if we could design this in a few weeks, any other team or company could do the same.
On the slide I also wrote down a limitation, because I knew that Faundit could add multimodal matching and Loggit could add PMS integration as well. If I am honest, I sometimes wondered if our combination was really something different or if we were just trying to save an idea we had become attached to. I do still think it can work for large hotels, because there the stay context and the verification matter the most, but I also know that we can only really find that out with a pilot.

The future-market test
For the future-market test we looked at a scenario where guests use AI agents to file their claims. While writing it I realised that I was actually arguing against my own Premium tier, because the time AI saves becomes smaller when claims already come in structured.
Akerlof (1970) helped me understand what was changing. The person who files a claim knows if the item is really theirs, but the hotel does not. When AI agents make filing a claim almost free, more dishonest claims will come in and it becomes hard for the hotel to tell them apart from the honest ones. Verification then becomes the bottleneck, which meant that our staff verification became more valuable instead of less, so we also added a measure for fraudulent claims to our test plan.
Eisenmann, Parker and Van Alstyne (2011) showed me a risk I had not really thought about before. A PMS vendor controls the guest data we depend on and could bundle its own recovery feature, and an OTA could do the same through its guest agent. Until then I had mostly been looking at startups like Loggit and Faundit, but this made me realise that the biggest threat could come from a company we actually need to work with.
This is the part where I got some of my confidence back. Our idea was still not unique, but I saw that it could still work if we were willing to change parts of it.

The team
After Loggit we decided as a team to stop comparing ourselves with competitors all the time and to focus on the value we deliver. That helped us to move on, and you can see it in the final recommendation to start with a pilot and only scale Premium if AI saves at least 3–5 minutes per claim.
Looking back I kept too many of my doubts to myself. In a next project I want to share them earlier, because if I can back up my doubts with sources, the team can actually do something with them. I also want to be more willing to let go of an idea when the evidence points in that direction.

Looking back
At the start of this assignment I mainly saw AI as something that makes it possible to build ideas quickly. After this project I see that this is also the problem, because if it is easy for us it is easy for everyone. For me the value is now more in the things that are hard to copy, like the data you collect, how you are connected to other systems and whether people can trust you.
So back to my question from the start, how do you still come up with valuable ideas when everyone has the same tools? I don’t have a full answer to that and I still have some doubts. What I do know is that in a next project I would look for competitors on the first day, before I get too attached to an idea, and that I would think about the future market earlier instead of at the end. I think the doubt I had is not something I need to get rid of, it actually helped me look more critically at our idea.

References
Akerlof, G. A. (1970). The market for “lemons”: Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500.
Barney, J. B., & Reeves, M. (2024). AI won’t give you a new sustainable advantage. Harvard Business Review.
Eisenmann, T., Parker, G., & Van Alstyne, M. W. (2011). Platform envelopment. Strategic Management Journal, 32(12), 1270–1285.
Rhodes, J. (2026, August 26). We deliberately left phones and wallets behind in UK hotel rooms. Which?. https://www.which.co.uk/news/article/we-deliberately-left-phones-and-wallets-behind-in-uk-hotel-rooms-aNNM24P3V54O