McGraw Hill, Inc. Goldman Sachs Communacopia + Technology Conference 2026

Thanks, Stuart. It is great to be here.

Yeah. I think it is super important to understand education. I think a lot of people have this misnomer, you just park a child or you park a student in front of a chatbot and you get a great outcome. The speech that we deliver is some of the most highly regulated and politicized speech on the planet. What you put in front of a child is legislated right down to the zip code level, the language you are allowed to use, the science you are allowed to use. We serve Saudi Arabia, we serve Oregon, we serve Texas, we serve Florida. It is extraordinarily, first and foremost, I will say a politicized environment, which creates a huge regulatory moat in the K-12 space. These adoption cycles tend to be long, 5-7 years for a lot of the core.

They are picking a partner to go along with the curriculum over a 5- to 7-year journey. Curriculum tends to be about $0.01-$0.02 of the education dollar. By that I mean, the largest cost is teachers. What you do not want is your teacher sitting up at night trying to make sure you are correcting a chatbot that has hallucinated or that they do not have a way to assess to see whether or not the child is actually grasping what is happening or not. In the K-12 space, really deep, I will call it regulatory moat. We are probably one of the largest curriculum providers in the world. We are in about 82% of all the school districts here in the U.S. We are one of the few that serves the Middle East.

A lot of the Middle Eastern countries, in some cases, are serving as much as 50% of their curriculum. In Latin America, a really strong position. As you get up into higher ed, there is a similar dynamic in the sense that the professor really wants to be able to have the trusted content to go along with what he is educating. In the medical space, we just were talking about what we were doing in the agentic space. You need to have medical-grade answers. I would like to say that in medicine, you need to be able to answer 99.9999% of the time with the exact right answer. That is extremely hard to get medical-grade AI, is what I will say. It is equally important when you are teaching engineering. It is equally important when you are teaching accounting. It is equally important when you are teaching mathematics. It is equally important when you are teaching literacy.

It is accuracy, quality of content, and then what I will say is the local pedagogy or the local requirements is important. AI for us is a huge opportunity for us. I like to talk about the fact that we have one of the largest, I think we can safely say we have the largest subscriber base to education, paid subscriber base. If you take a look at something like Duolingo, they have about 12.5 million paid subscribers. I think Khan Academy I saw might have 35 million. We have over 100 million paid subscribers to our content. Our goal is to expand the number of students, the number of subjects we teach, and we think that there is just no end to the things you can teach for humanity, and it will not be. Also the way that you teach.

We have this strategy to be able to build AI tooling on top of the piece of the curriculum. When you build agentic tooling, the four things that you must have in the agentic world is that you have to have content, you have to have context, you have to have an LLM, and you have to have tooling to make sure the whole thing works together. You will see us, we are delivering some of those things today in packages like Sharpen or ALEKS that can help teach a student algebra or be able to do prompting at night for the student. You are also going to see us build an additional set of agentic services on top of each of those 100 million licenses.

First of all, one of the things is trust is essential. What you just saw this past week where both L.A. and N.Y. actually banned AI chatbots in the classroom. In L.A.'s case, it's K-12. In N.Y., it's K through 9. There's not a lot of trust. We just did a survey, it said teachers, about a couple of thousand teachers that we surveyed said 81% will trust AI if it's embedded inside of the tooling that has already been improved. Outcomes matter, first and foremost. I've talked a lot about the fact that the average school has over 2,400 learning tools, and we're not getting the outcomes. Increasingly, proof of outcomes in the pedagogy really, really matters. I've got to always start with that.

Trust is the most important thing in implementing AI in education, and we start from a pretty amazing position. The second is in this area of what I call the ontology of learning. When you teach a child literacy, you're teaching somewhere in the order of about 3,000 skills from kindergarten all the way up to 12th grade. You literally, I can tell you in the fifth grade on the third week what the activity is going to be for that student. You literally lay down that fine of a grain, and you have thousands of these learning objectives that you have to layer on top of each other. When you teach something like Algebra 1, you're teaching 400 skills in a particular semester or class. When you teach Biology, it's about 4,000 skills.

When you go into medicine, my son just graduated medical school, and I'd like to say that he graduated with 54,000 Anki note cards. We think that there's about 54,000 things that you teach in medical school. Is the student on track or off track? How does a chatbot know that? We have this open learning record, and because we create the content, we also know what these learning objectives are. You need to be able to have a source of proficiency, a true understanding of proficiency. Whether or not you are understanding 100 of the concepts in Algebra 1, or whether or not you're understanding all 300, and what's your proficiency level, that's the second thing.

So we've got a tool like ALEKS that each learner has about 200 to 400 states when you're learning algebra, and with just 25 questions, we can tell you precisely what that learning proficiency is. So the ability to be able to have data and say are they on track or are they off track. Students that are using ALEKS, they're able to achieve much higher levels of proficiency much more rapidly. Then we have a tool like Sharpen as an example, and Sharpen is this great example. Think of it as Spotify for the professor. They can push playlists out to the students, and they're able to engage with that content at all hours of the night. Most of them are engaging, believe it or not, between 11:00 at night and 3:00 in the morning, along with all of our tools.

That ability to be able to tie it back into the professor, because the next morning when the professor wakes up, they can see what kind of engagement there was, who's overly confident in their answers, and what are they searching for, and what should they be teaching. So tying it back to the teacher and giving the teacher the ability to be able to guide the classroom is really, really important.