Introducing All Day TA
My new venture designed to bring AI responsibly into college education
This is an exciting day. For most of the past two years, my Rotman colleague Kevin Bryan and I have been experimenting with using generative AI to improve our students' education. What has emerged from that is an AI teaching system that takes a professor’s content and creates a way for students to ask questions about their course and get reliable and informative answers all day, every day. We were so happy with the results that we bit the bullet, put up our own money and time, employed some real people and launched our own AI startup called All Day TA.
I could explain the venture in words, but it is easier to show you, as I do in this video.
You might be thinking a few things coming from this. I hope you are thinking, “this is neat,” because that is what we are going for. But you are also thinking, “aren’t there likely to be a ton of people doing this type of thing?” A very natural thought. We are economists, so why enter the fray here?
As it turns out, there are plenty of reasons. First of all, other free tools like making your own GPT using OpenAI’s system or using something like NotebookLM are fine at a first pass, but aren’t tailored for education. They provide average responses for lots of uses. By contrast, with our system, we take the content, recognise its educational value and push answers to students based on that intent. In other words, it is less standardised and more personalised to what professors want to achieve.
Second, this is software, and a big potential market exists. Won’t big companies with big budgets invest millions of dollars in making courses that really can engage students with multimedia and the like the Khan Academy does? Yes, that is true. But which courses? If you are thinking anything from elementary to high school, there are big markets for each course, so that investment is worthwhile precisely because many students are learning the same stuff in a standardised way. This might even be the case for larger first-year college courses too. But the vast majority of college courses are not like that. They are small and specialised, often with 50 students or less, and are taught in idiosyncratic ways according to a professor’s judgment and expertise. The answers a professor wants their students to receive are not those that Google or ChatGPT can reliably provide. The goal is to move students from those sources but, at the same time, allow the advantages of having an automated assistant.
Third, but even so, won't there be many startups that can provide this type of service? The answer is yes, there will be. Competition is coming. I wouldn’t be surprised if many engineering students (or their professors) have had this idea, and some may well be doing it. Here’s what’s different about what we are doing. I have been innovating using online tools for education for almost three decades now. This has been done with tiny budgets, and, in most cases, the tools were then developed by larger companies like Blackboard and Canvas, who then provide tools that are fine but also, frankly, still hard to use. AI TAs are coming and will be with us. This time, I wanted to get ahead of the game, take it very seriously, and do it right. I hope you saw from the video how easy it is for a professor to set up and use the system. That is by design. We worked through every click and created a flow to make it as easy as possible to upload content and distribute it to students. I want professors to love this product. That’s the goal. I believe it will be what differentiates us from the competition.
If you are a professor, you can sign up and try All Day TA today. It is free for two weeks and then costs just over $1 per student. (Alas, AI has costs.) During the Fall, about 5,000 students used it in a pilot at Princeton, UCLA, Berkeley, MIT, UNSW, Melbourne, Monash, and Toronto. On average, a student asks 50 questions a course. So that is not simply saving a professor some emails but students are able to ask questions where even they didn’t think an email was worthwhile. In the Spring, we will be implementing student self-testing and other features. We are excited about the current product but even more excited about where this might go in the future.
For anyone else, here is a link to a TA I created to demo AI Strategy based on my books, articles and this substack. Feel free to try it out and send me any comments you might have.
Finally, if you are curious about our overall philosophy regarding AI in education, here is a video webinar Kevin did that describes that in great detail.