“It is not easy to gauge whether students are with you or struggling.”
This was one of the responses we received when we emailed professors about how they notice students are struggling. This statement was exactly the problem we wanted to address with AskCanvas: an AI assistant integrated into Canvas that answers student questions and gives instructors an overview of recurring misconceptions.
When we started the AI Strategy Lab, I was thinking too much about what AI could make possible, instead of how it could turn into a business strategy. I assumed that if we developed something useful that solved a real problem, the advantage would follow by itself. But as we worked on AskCanvas, I started to realise that creating value and capturing it are two different things.
After submitting our proposal I thought it was quite convincing. However, after reading the feedback, I realised what we, by then did not understand. We had explained quite well how AskCanvas could create value, but we had not really shown how Instructure (the company behind Canvas) could capture that value.
This was especially relevant to my part of the project, since I was responsible for the economics. I also worked on the final presentation with a teammate, where we tried to bring everyone’s work together into one clear story.
One thing I struggled with was putting a number on the value AskCanvas could create. I mainly focused on how much time instructors could save, because this was the easiest benefit to measure. However, my calculation assumed that every question answered by the AI would otherwise have been answered by an instructor. If only a quarter of those questions would actually have reached an instructor, the estimated savings would be three quarters lower. I also realised that saving time does not automatically mean saving money for universities. Instructors would still receive the same salary but could spend that time on other teaching activities instead. Other benefits, such as improving teaching, were much harder to measure. Looking back, I think I was too focused on finding numbers that would make our idea convincing. I wanted AskCanvas to work, so I did not question some of my assumptions as much as I should have. I still think it was a good starting point, but I should have been way more aware of my own assumptions.
Initially, we considered charging per question, but it didn’t sit right with me. The dashboard depends on students asking questions, since these reveal where they are struggling. If universities had to pay for every question, they might want to limit usage. That would mean fewer questions and therefore less information for the dashboard. We would basically discourage the thing that makes AskCanvas valuable. My teammates agreed, and we started looking at other options. This was one of the moments where something from the lectures really clicked for me. Pricing is not only about generating some revenue, but also about how the way you charge affects the value of your product.
Our original idea was to include it in one of Canvas’s existing subscription tiers. But, after discussing this, we realised that universities that were not interested in the feature would then also have to pay more. We therefore decided to offer it as a separate add-on with Base and Premium versions. I already knew about versioning from the lectures, but actually applying it was more difficult than I expected. I found it interesting that something which seemed quite straightforward during the lectures was so hard implementing into the project.
Since I had been working full-time for the months before starting BIM, it took some time to get used working on a university group project again. But it also reminded me what I like during group work. I prefer having my own responsibilities but still sitting together so we can easily discuss things and help each other. What I found difficult was how long it took us to choose and then develop an idea. In the beginning, we kept discussing different options without really making a decision. After receiving the feedback, we became much more focused and worked well together. Personally, next time I would like to take more initiative in making a decision, rather than participating in the discussion continue for too long.
During the last few days before the deadline, we sat together and went through everyone’s work to make sure the whole story was coherent. That was when I really noticed how much my economics depended on what the others had done. The prototype and the test showed what we could actually claim about AskCanvas, and my calculations had to make sense alongside those findings.
The project also changed how I think about AI. Before this course, I was thinking AI could be the best for automating and thought the technology itself could give a company a competitive advantage. But as we discussed in the lectures, competitors can often use the same models. Through AskCanvas, I started to see that the advantage comes more from how the AI fits into a platform people already use. For Canvas, that is also why a feature like this matters. If universities start using AI tools outside Canvas, the platform could become less central.
I am happy with what we achieved. We managed to develop and test an idea that addressed a real problem, and we improved it based on the feedback we received. The biggest lesson for me was that having a good idea is only the beginning. You also need to understand who benefits from it, who is willing to pay for it and whether the assumptions behind your business model actually hold.
