A brief introduction
For AI Strategy Lab our team designed ReUnite, an AI-assisted lost-and-found platform for hotels. Staff register found items, guests submit reports, and the platform brings reports, items and analytics together in one place. On top of this, AI matching links guest reports to the items staff have registered and ranks the most likely matches, so staff can verify them faster.
My main contributions to the project were the economics, the prototype and the evaluation. In the economics, I estimated the staff time saved per case and the costs of AI usage, and helped shape the pricing model. I built the prototype, which turned our idea into something we could click through and try out. In the evaluation, I helped design how to test our core assumption: whether AI matching saves staff enough time, compared with a well-organised digital registry, to justify a premium price. These three parts depended on each other, since the prototype determined what we could test and the evaluation determined which economic assumptions we could support.
The role of AI in the business strategy
I learned several things about using AI in business processes. One of the main traps of AI nowadays is that there is too much focus on AI itself, rather than on the problems it should solve. This could lead to for example making an AI tool without having a problem it should be able to solve, and thus ending up with an AI tool that doesn’t have a lot of use. Another problem that could occur is creating an AI solution for a problem where a different, maybe much easier solution, would’ve sufficed.
For our project we made a product that includes AI, but where it isn’t necessarily the main solution. It isn’t supposed to replace humans, but rather help them with their work. I have learned about this during one of my main contributions to this project: the prototype. During the making and testing of our prototype I found out that the product we made would be useful on its own, even without AI, because it was an organised and centralised place where everything regarding lost-and-found was available. Our prototype didn’t have all the possibilities included yet, because of the scale of the project, but it included already things like guest reports, staff reports, analytics and an overview of items. So the fact that this all was included in one place already made handling lost-and-found requests a lot better.
There was still room for development next to this, with the help of AI. Because with the help of AI with matching of photos of guests and staff reports it would make the handling of lost-and-found requests even better, even though it still required hotel employees to work on it themselves. But if hotels have a lot of items in their lost and found the help of AI matching to show the best matches can help even more.
The problem that caused a change
Before starting, we knew that companies were already active in the lost-and-found market, but we did not see this as a major problem. We submitted a short proposal, received approval and started working. It was only during the project, as we researched more deeply, that we found how closely some existing solutions, especially Loggit, already resembled our idea. Faundit, Loggit and Deliverback all already offer digital registration, and Loggit even offers AI photo matching. We had assumed that AI matching was more distinctive than it turned out to be.
This forced us to rethink two things. First, we could not present AI matching as our unique selling point, so we had to find other ways to stand out, such as combining matching with hotel-specific context, ownership verification and the return process. Second, it changed my economics work. Competitor prices were low (Loggit, for example, charges £35 per venue per month), so a flat subscription justified only by AI would face strong buyer bargaining power. We moved to a tiered model with a basic and a premium tier, plus a guest-paid return fee.
We still found ways to differentiate ReUnite. Our matching uses hotel-specific context from the property management system, such as stay dates, room and outlet, and not only the item and the guest’s description. Our Premium tier combines this AI matching with PMS integration, coverage of multiple outlets and analytics, while the Basic tier stays close to the market price. Staff also verify ownership before an item is returned, and verified matches can over time become data that improves the matching.
The lesson for me is that market evidence should be gathered early and repeatedly, not once. Our first check was not thorough enough to show how similar existing solutions already were, and the real picture only appeared once we looked closely. We had to change our direction a bit, but we still believed the idea had potential, and thinking harder about differentiation made the proposal more realistic.
Collaboration with a change
During the project, one of our four members dropped the course and therefore the project. All groups had already been formed, so we had to continue with three people. That meant the same work had to be divided among fewer people. The collaboration still went well. Our different schedules sometimes made it hard to find a time when everyone was available, but once we met, our meetings were clear and structured. Everyone took responsibility for their own sections, and we supported each other when it was necessary.
What I take away
This project changed how I look at AI in business. I learned to start from the problem, to test the simplest solution first, and to treat AI as a tool that must earn its place. I also learned that AI alone rarely creates a lasting advantage, especially when competitors can offer similar features, so value has to come from elsewhere: the workflow, the data, the trust and the integration around it. On a personal level, building the prototype and working on the economics and evaluation taught me to connect technical choices to business value.
