The Messy Middle of AI in Education: Panic, Promise, and What Comes Next

0:00If we can give educators real-time feedback, if we can give them the right screening and progress monitoring tools to understand what's happening with this learner, and we can turn educators into more agile, personal supports to these students, I think we will see a shift.

0:25Matt Paige:Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters. I'm your host, Matt Page, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI. Every generation of educators faces a transformational technology. We had calculators, the internet, Google, now generative AI, and the pattern's always the same. A resistance, then grudging adoption, then a realization that the tool doesn't replace learning. It raises the floor for what's possible. And right now students are using ChatGPT among many other AI tools to shortcut homework while teachers scramble to figure out how to respond.

1:05Matt Paige:And the real question isn't how to stop them. It's how to redesign the system. So the technology actually makes learning deeper, more personalized, more equitable. And Todd Brekus has been building ed tech for 25 plus years from the early days of the internet in classrooms to co-founding and the president of Mayon to now leading product at Renaissance, one of the largest and most data-rich ed tech platforms in the world. They just launched Renaissance Intelligence, their first AI native education intelligence system. And today we're getting into what it actually takes to bring AI into the classroom in a way that empowers teachers, meets students where they are, and doesn't just add another disconnected tool to the pile.

1:47Matt Paige:But welcome to Talking AI, Todd. All right, Matt, thank you. It's a pleasure to be here. Yeah, this is going to be a good one. I've been wanting to have somebody in the ed tech space for a very long time. I've talked about it with several guests. I'm excited to have somebody that's actually deep in this space. But Todd, I want to start with something I saw firsthand that kind of crystallized this whole topic for me. And I've mentioned this before in previous episodes, but I take my daughter to dance every week and the studio is too far away from my house to drive back home. So I ended up working at a Starbucks nearby and it's right next to a high school.

2:23Matt Paige:So there are always kids in there. And one day I was sitting at one of the communal tables and I couldn't help but overhear these high school students go back and forth with ChatGPT about their homework, just out in the open, totally normal to them. And I couldn't help myself. I had to ask them, how often do you use ChatGPT for your homework? And I'm not joking, Like without hesitation, every single one of them was like a hundred percent of the time. Like didn't even cross their mind to think about it. And the thing that hits me is, you know, the teacher's assigning the homework, the students pumping into the chat GBT, chat GBT is giving them the answer.

2:59Matt Paige:Then they give it back to the teacher. And let's be honest, the teacher's probably leveraging AI to help build the lesson plan and in some cases grade it. So you've got AI on both sides and the actual learning is getting lost somewhere in the middle. school you've been at ed tech for 25 plus years you were in classrooms when schools were literally figuring out how to use the internet on the ed tech space when you see something like that what's your gut reaction are we in a moment of panic or is this just that messy middle of something that's actually going to be really good for education oh it's such a good question and first of all i can reflect on the dance, uh, the dance, a dad point.

3:43I don't know if you've had to get in your outfits yet and do the daddy daughter dance thing yet. Okay. Yeah. Been there, done that. Uh, my daughter's now a senior in college, so has lived through both the dance pieces as well as the, the chat GPT world of, of, of education. So yeah.

4:02Matt Paige:So mine are 10, 10 and a four-year-old. What's how I know it goes by quick but like literally from 10 to college what did that feel like was that just like snap of the fan oh at the time no but now it's like oh my gosh she's graduated from college this year and it's like wow it's it's it has gone very very fast and absorb and enjoy every moment and all the embarrassment she can pour on you the better yeah i love it it's already derailed there but she just hit double digits and i'm like man this is very fast so yeah i have four four kids only one dancer but that was my last child and she was uh she was amazing at dance and it was so much fun but a lot of days in the parking lot wait for dance pickup them yeah but this moment that's going on yeah like is it which what's your take on it because i know a lot of people have varying reactions of good, bad, weird, you know, which, what's your thoughts on it?

5:02Well, I mean, I think the first reaction is the kids got there first. Like, uh, my son, November, I think it was Thanksgiving break. He was a, he was a junior in college, came home. There's this new launch that just happened this week and showed me Chad GPT and it was incredible. I was blown away. And yeah, of course, every paper he did for the next year and a half was completely using some type of agent, some type of bot to help him support his writing. And we talked a lot about what are your ideas? What were your ideas versus what were the systems ideas? And did you understand? Quick break in the pod.

5:46Our State of AI

5:47Matt Paige:2026 report just dropped and it breaks down what actually is changing in AI, what's hype and what leaders need to be paying attention to this year. You can grab it right now on our show notes or at hatchworks.com. And now I'm a former educator, so I'm not going to let him just throw something back to the teacher without knowing it himself. And, and, but not all parents have that time or that ability. And, and frankly, the educators got, got a little bit caught flat footed on this model. And it's, I think, a little different than the internet in that, so I got a backstory myself. I was teaching teachers how to use the internet.

6:24That was my first job after teaching. I was fortunate enough to travel around with Dr. Vince Cerf, the creator of TCPIP and the internet, and got to follow in his footsteps after he'd go give presentations. And then I would help the schools get wired. I would help teach the teachers how to use the internet. and back at MCI, MCI WorldCom, which is now Verizon, was a really powerful start for right out of teaching and learning. And we trained about 50 ,000 educators in that first two years, and it was incredible. But in many ways, we were, you know, the model of the Internet was, I can find something.

7:04Like I can remember the first time I looked at the Louvre, and, you know, you got the Mona Lisa coming, and it took only a minute and a half to actually get that picture on the screen. And then you'd go, imagine what's going to change with education if you can see all these different things you haven't been able to see before. And it was an aha moment of the information is in front of me. It's here. I don't have to go to that physical library. I don't have to pull out that research text. Look how much easier this is. And so I think the easiness factor was there. The data model of information is available to us and we should consume it.

7:44That's always been part of education. I think this model of generative, that something is getting created by a machine, it changes the game quite a bit. And I don't think educators are quite ready in some respects. I think we've gone from you've got to buy a product to make sure cheating hasn't happened, and now you've got to buy a product to make sure AI hasn't done it all for you. Now there are systems and tools that the colleges are putting in to make sure that copy-paste is tracked and how that is working. And yeah, granted, but the worry I have with the generative AI piece is in many ways, we're giving it to students too quickly compared to what we're giving to teachers.

8:31And I think we need a significant program of educator awareness, educator understanding. And then our top officials need to start thinking about what does curriculum and instruction really look like for the future? and how do we incorporate an AI-driven model into teaching and learning so that we, the smart adults who know how to deliver education to kids, can do it in a mindful way that has the technology at hand and uses it effectively but doesn't just fill in the answers. So teaching and learning has to change quite a bit to adapt to AI. And I think that's such a good point there.

9:13Matt Paige:And in my opinion, I'm glad the students are pushing this. And to be honest, I think I envisioned this game of whack-a-mole where it's like, oh, let's, we got to detect AI. And I'll be honest, I see tools out there that are like, okay, you're going to try to find the copy paste. I'm going to build a tool that looks like it's naturally writing something out. So it's just going to keep happening. Or I'm going to write, I'm going to write the prompt in such a way that I'm going to say, don't make it look like you actually were, make it look more like me. here's my last 10 writings and become my own agent, if you will.

9:48Matt Paige:100%. And the models are getting so good that it can do that. And it reminds me of, like you mentioned, I always feel like I'm young, but I'm about to hit 40. But when I was in, I can't remember if it was middle school or early high school, Google was like the new thing, right? And before when we did a research paper, we had to learn the process of going to the library. How do you check out a book? How do you research and find all the detail you need? And then we had Google and we started doing things differently to where we would go to Google to do our research. And it completely circumvented that whole process before.

10:26Matt Paige:And there was, I think, you know, a lot of resistance to that. But what was enabled was I could do way more, way more meaningful work with this new tool. So I think the interesting part was, okay, how do educators adjust to the lesson, the approach, what they're asking of students now that they have this new tool in their hands? And I'm curious too, you were president at Mayon when the idea that books don't have to be restricted to a place in time, unlimited digital learning, which sounds very obvious now, right? But it wasn't then. what gave you that conviction that that would work and that would be a thing because i think we're seeing similar patterns now where this feels like some weird thing with generative ai but i think several years down the road it's going to be obvious that we're leveraging this yeah i boy there's so much to get into and i i'll and first answer your question about i think kind of connecting the dots of the internet and the usefulness of the internet was if i'm waiting a minute and a half for the louvre to show up and all of a sudden wow i've got a painting in my room I'm like, ah, it doesn't really change teaching and learning.

11:36Maybe I get an answer quickly about the War of 1812, or maybe I get a specific link to other research and information that can help fill in my gaps. But I think that we absolutely had to shift education to be much more of how am I articulating knowledge? How am I articulating information? How am I synthesizing this information? How am I proving fact versus fiction? I think that was the difficulty in teaching and transformation with the teaching of the internet. Now we're actually talking about how do I produce something and create something and show knowledge when I'm giving this over to an agent, to a generative bot?

12:24And how do I ensure that I am not getting less smart and I'm actually increasing my information because i think what what the the internet age gave us is hey i have more information quickly i can synthesize more i can speed up some of the break you know the difficulty of going to get all that information and going to the research library and going to this place and going to that place to get the information and so the transformation with myon was look we should get all these books digitized. We should have this ability. Why do we have physical libraries just to check in and check out? Remember, probably when you were in middle school, you went to check in your book.

13:09You couldn't check in the last one because you didn't turn in the two books before and you were restricted from reading. That doesn't make sense. We're in the age of on-demand video, on-demand television, on-demand everything. Why can't we have on-demand books? And the challenge at that time was we didn't have the delivery devices in the schools and the level of licensing models that worked for schools. So just like Netflix came about to provide us an unlimited on-demand movie model, we didn't have that for education because schools weren't used to purchasing an unlimited on-demand SaaS model for reading and literacy.

13:46And that was the simple, the transformation there was a business model change where schools didn't have to buy every single book, just like we don't have to buy a first right to every single movie. We get all of these movies and we can pick the ones we want to go to at the time. So frankly, the video companies, especially Netflix, were well ahead in this model on the CD-ROM side. And when it went online and we were seeing rumors that the iPad was coming out and this physical device, we'd already had the Nook. And that was a business model change. And at the same time, my passion was I want literacy to become unlimited.

14:27I don't want it to be restricted for kids who, you know, if the schools are going to give these devices and they're going to do the Chromebook activity and they're going to do the rain iPads in the classroom, I want kids to have unlimited reading opportunities and not be restricted. And so that it was a it was a creative business model. And it was the publishers ready to adapt to a new business model, similar to the movie production studios that had to adapt Netflix model. They had to adapt in an unlimited reading model. And we were really the only ones at the time to have an unlimited subscription model reading, unlimited environment, and then to create the connections to the classroom.

15:04What do teachers need to do to transform? If you have an unlimited nonfiction library in the classroom, you don't, everybody doesn't have to pick one book. You can pick multiple books. You can think about the biographies you're going to read or the autobiographies, et cetera. So anyway, the Maya model, I think the transformation there was a business model, but also a shift in, wow, if I have all these resources, I don't have to wait for the book to be checked in or checked out.

15:30Matt Paige:Yeah, I get chills kind of thinking about that a bit because in my mind, I'm thinking how that applies in this new age. And this is all very personal to me right now because I'm going through it firsthand as a parent, kids in school. And so my daughter struggles with learning a bit. She gets outside tutoring and help like that. And it's just how her brain works, right? She processes things differently, super smart in her own way, but sometimes the traditional method doesn't work. But this idea of personalized learning, similar to how any kid can go read anything, being able to have some personalized nature that's tailored to how she learns just blows my mind in terms of the potential.

16:15Matt Paige:And then you look at underserved communities and not just like here in the U.S., but globally. Right. So anybody can literally have effectively this teacher, you know, through generative A.I. that can help them progress and learn and just spark their imagination. And to me, I think that's the power and the potential in it, but it's right now square peg round hole very much, you know, in terms of like the current education system. But the personalized nature of it, I think that's in my mind where a lot of the potential lives. And I feel like there's parallels to what you just talked about. I'm curious your thoughts on the personalized.

16:57Yeah, I think everybody can see the leap to the personalization of the student. And before I go there, I think my bias right now at this season, at this timeframe in education technology, I think we need that personalization to get to teachers first. And I'm not saying that kids shouldn't use the AI. I'm not. Yeah, of course, use the technology. I've always been a fan of using the technology, but my worry is that the teacher's opportunity to really understand these young humans that are in our classes that have this tremendous potential, I will strongly believe that a human interaction and a human connection, I think we can all remember our favorite teacher, our favorite breakthrough in a subject area, our favorite, you know, when we persisted and we tried so hard and we finally got it.

17:48But that aha moment is, I think, very hard to do in a simulated environment compared to a human environment. And I think there's a balance between where the simulated versus where the human side of this is. And what I'm really focused on right now and what our company is really focused on right now is how do we give the best tools to the teacher to personalize? And so while personalization sounds like it needs to get to the individual student, and it does, at the end of the day, the learner is the one that's got to pick this up. That's got to start to grow. That's got to build that self-efficacious model of, I want to learn.

18:29I want to grow. I want to be successful. I want to please mom and dad. I want to do the right thing. They've got to start to take that ownership. And that will happen. But all of these foundational skills, all of these, especially in the young ages of your children right now, that educator is critical to this guidance in the model. And so for me, this is how do we personalize the teaching and learning? And so if we get the power into the hands of the teachers in a way, like let's just take literacy acquisition for a moment. We know that students, we know the science says about literacy acquisition.

19:10And so if a student has missed some of those steps in those stages and we don't, and the teacher doesn't have any data about what is missed for that child, they're going to just teach the whole class the same way we were taught. Like, okay, everybody, everybody repeat after me. And I'm sorry, some of those classes can be a real, real bore and frankly, a turnoff to learning, especially when I got all these cool tools at home that can just tell me the answer. why do i have to go learn the answer and why do i have to learn the methodology to get towards how the code comes together to read or how to how a sentence makes uh meaning uh and how a meaning of a sentence makes a paragraph and a paragraph makes you know all the components of teaching the bots that we've died to do with generative ai and all those interconnections of subjects and verbs and all the rest but i guess the point being here if we can give educators real-time feedback, if we can give them the right screening and progress monitoring tools to understand what's happening with this learner, and we can turn educators into more agile, personal supports to these students, I think we will see a shift in, you know, this, everyone learns together in this one big environment.

20:34Everyone learns at the same pace at the same time. And that's, I think, has shown us in the last 60 years, we have not seen any growth in NAEP scores or in any scores in general. And since COVID, we've obviously come down. So the question to me is, how do we help the educator personalize teaching and learning? And I think data is a big piece. I think we're all focused on the generative AI piece. And I think, obviously, it's an immense breakthrough and it's going to transform and change. And I believe all of the things that, you know, whether it's Sam Altman or Elon Musk or others have said that I think it is absolutely going to be changing how information and data and programming and creation and optimizations are going to happen throughout our industries.

21:21But for this little young learner that's maybe six years old learning how to read and that teacher, how can that teacher personalize for this six-year-old, this eight-year-old, this nine-year-old in a way that they have the data to inform them, they have agents and support tools available to them, and they can deploy AI in a way that affects that individual learner. those groups of learners having similar experiences and in a sense manage multiple kids in a class at the same time at different paces and spaces yeah that's that i think is a really interesting one

22:01Matt Paige:that's that's really smart because at the end of the day like young kids they're worried about what's for lunch what am i doing at recess today they don't know how to be their own advocates it's almost teaching the that's right the teacher how to fish in a sense because once you empower them and turn that light on for the educator, like then they, they're the biggest advocate for the children. So that actually makes a lot of sense. Cause if, you know, it's almost the wild west in, in to your point, the educator can be the biggest bottleneck in the system. If you don't win them over on how to leverage this new technology.

22:32Matt Paige:I will say though, my daughter was learning conjunctions the other day and I still went back to conjunction, junctions, what's my function. Exactly. Yeah. There's still some stuff from the old school that still works very well. But you touched on data a few times. And like, so obviously we're, we help, you know, companies and enterprises with their AI adoption strategy, all of these things, but it all comes back to data at the end of the day, even with AI, it's at the foundation. But I think a unique thing that you're doing with Renaissance Intelligence is the trusted data element, which, you know, you want to make sure the stuff that you're leveraging is truth.

23:12Matt Paige:And in fact, and all of these types of things. How are you thinking about the data elements and whether it's the accuracy and all of those things related to it when you think about this in the system that you're building with Renaissance Intelligence? Yeah, I mean, I think if you think about education today in public schools, I'll pick on public schools. The majority of our goal in the public school system, obviously, it's that everybody makes progress. Everybody grows. Everybody matriculates with grade level expectations being met and those standards that happen every single year. But the challenge is the measurement of that is the end of your test for the state.

24:00And it's a measure of the school. It's a measure of the school district. It's a measure of all the subgroups underneath that. But it's in May. It's in May for the most part. Florida, they're doing this three times a year. And Nebraska, they're doing maybe three times a year as well. But they're waiting till the end of the year. Did it work? And so the data in general is incredibly latent. It's incredibly, it's very waterfall-ish in the software world that did we complete our release? Did we finish everything? Did we get it all done? Did we actually score in our proficiency range that we expected to be in?

24:43And that is a big problem if we're not driving the information to the moment of instruction or the moment of teaching and learning or within that week or the month or at least the quarter. How do we not let a whole year go by and not adjust for teaching and learning? So I don't think that sounds like 1800s, but it actually is most of our states today. assessments are happening the main measure of that progress for that school which the federal government requires is that students make adequate purely progress um that's passed into law in the program within our federal government now granted states have bought products like renaissance and a couple of my other competitors so they can do three times a year assessment so a beginning year, middle of the year, end of year assessment, and then compute growth out of those three tests.

25:43I'm going to argue that's not enough either. And I think that even if I move from waterfall software development and I, you know, my software development life cycle goes, oh, I'm going to, I'm going to do OKRs and I'm going to measure it per quarter. But I think anybody who does, and I'm going to end up in the software world because I'm speaking to someone who's been in good analogy for your space is now we have continuous development. Now we've got GitHub and we've got these tools that I can do continuous development. I can test these things. I can launch darkly, or I can do these, these AB tests very quickly to see what's working.

26:20And we see companies that have massively transformed the SaaS industry. And obviously AI coming to this mix. If you don't have the data, you don't have the measures, you're not going to prove if the AI actually works. So if you think about this in the education space, how do we get more exhaust data to understand if it's working faster so that we're not getting to your mid-year review with your child or an IEP review that happens every quarter or some type of we've helped them. We got to measure quickly and we got to measure at chunks of the information to see if it's working. And so I'm more of the mindset that we need a more mastery model in education where we see a student progress with something, they master it, they can prove that they've mastered it, we move on, we skip the things that they don't actually need, we adapt the classroom somewhat.

27:14Granted, I think standards in general have to be covered. We need to cover the curriculum. We need to cover what we as a state or what we as a country have said, these are the requirements for all learners to learn and everybody should learn these things. But maybe we can personalize that instructional pace if we measure more frequently and we measure with speed and accuracy and it becomes part of an instructional approach, not just an assessment and then an instruction approach. And so I think at the main point here is there's a measurement model and we're measuring not these long, incredible tests, but in very short cycle, very short, periodic ways and make more agility happen in the educational system.