AI Should Support Teachers, Not Replace Them with Joy Delizo-Osborne of Student Achievement Partners

Joy Delizo-Osborne is President and CEO of Student Achievement Partners, where she works to improve instructional quality and connect evidence-based practices with curriculum and edtech. A former teacher, instructional coach, and principal, Joy brings more than 20 years of experience in education.

💡 5 Things You’ll Learn in This Episode:

1. How AI can support and not replace teachers and students.
2. Why productive struggle still matters in AI-powered learning.
3. What districts should consider when evaluating AI tools.
4. How edtech can better fit into existing classroom practices.
5. Why student agency should remain at the center of AI use.

✨ Episode Highlights:
[00:02:40] Student Achievement Partners’ role in improving instructional quality
[00:07:37] Maintaining rigor and quality with AI-generated content
[00:18:25] Closing the gap between edtech tools and real-world outcomes
[00:24:27] Why edtech companies need to understand classroom realities
[00:29:07] Building AI around existing curriculum and teacher practice
[00:40:36] When should AI enter the learning process?
[00:47:06] Putting student agency at the center of AI

😎 Stay updated with Edtech Insiders! 

Follow us on our podcast, newsletter & LinkedIn here.

🎉 Presenting Sponsor/s:

Innovation in preK to gray learning is powered by exceptional people. For over 15 years, EdTech companies of all sizes and stages have trusted HireEducation to find the talent that drives impact. When specific skills and experiences are mission-critical, HireEducation is a partner that delivers. Offering permanent, fractional, and executive recruitment, HireEducation knows the go-to-market talent you need. Learn more at HireEdu.com.

This season of Edtech Insiders is brought to you by Cooley LLP. Cooley is the go-to law firm for education and edtech innovators, offering industry-informed counsel across the 'pre-K to gray' spectrum. With a multidisciplinary approach and a powerful edtech ecosystem, Cooley helps shape the future of education.

Tuck Advisors was founded by entrepreneurs who built and sold their own companies, frustrated by other m and a firms, they created the one they wished they could have hired but couldn't find. One who understands what matters to founders and whose North Star KPI is the percentage of deals closed. If you're thinking of selling your ed tech company or buying one contact Tuck advisors now.

Every year, K-12 districts and higher ed institutions spend over half a trillion dollars—but most sales teams miss the signals. Starbridge tracks early signs like board minutes, budget drafts, and strategic plans, then helps you turn them into personalized outreach—fast. Win the deal before it hits the RFP stage. That’s how top edtech teams stay ahead.

Transcript

[00:00:00] Alex Sarlin: Innovation in pre-K to grade learning is powered by exceptional people. For over 15 years, edtech companies of all sizes and stages have trusted higher education to find the talent that drives impact. When specific skills and experiences are mission-critical, higher education is a partner that delivers.

Offering permanent, fractional, and executive recruitment, higher education knows the go-to-market talent you need. Learn more at hireedu.com. That's H-I-R-E-E-D-U dot com. 

[00:00:34] Joy Delizo-Osborne: So the fact that what comes out of AI is just procedural doesn't matter. When we start to think about an AI program as the full teaching experience or learning experience for a student, I think that's where things start to break, right?

Just how well is it tuned to do all of the things that a teacher with lots of experience, tenure, and knowledge of just like kids would do.

[00:00:59] Alex Sarlin: Welcome to Edtech Insiders, the top podcast covering the education technology industry. From funding rounds, to impact, to AI developments across early childhood, K-12, higher ed, and work, you'll find it all here at Edtech Insiders. 

[00:01:15] Ben Kornell: Remember to subscribe to the pod, check out our newsletter, and also our event calendar.

And to go deeper, check out Edtech Insiders Plus, where you can get premium content, access to our WhatsApp channel, early access to events, and back-channel insights from Alex and Ben. Hope you enjoy today's pod.

[00:01:39] Alex Sarlin: Welcome to Edtech Insiders. We have a fantastic guest today, somebody who's been really deep in understanding implementation and understanding standards, and really making sure that instructional quality is high across the board in schools and districts all over the country. We are speaking with Joy Osborne, the president and CEO of Student Achievement Partners.

She's a former teacher, coach, and principal in the Bronx, Harlem, and the Bay Area, and she leads with a commitment to seeing and valuing every student. She holds degrees from Pomona College and Claremont Graduate University. Joy Osborne, welcome to Edtech Insiders. 

[00:02:16] Joy Delizo-Osborne: Thanks, Alex. Thanks for having me. 

[00:02:18] Alex Sarlin: I am so excited to speak to you.

I-- You know, Student Achievement Partners is a really fascinating organization, and it-- I think it, it plays a really important role in actually making sure technology and standards are actually making a difference in classrooms. Uh, give us an overview of what Student Achievement Partners is and what your work looks like on the ground.

[00:02:40] Joy Delizo-Osborne: Totally. Thank you. SAP is an amazing organization. I've been a part of it for the last seven years, but it was actually started back in 2011 by the writers of the Common Core State Standards. So the primary writers of the literacy and math standards K-12 came together, and as they thought about how they wanted to shift and improve instruction across this nation, and then now even globally, truly, they wanted to bring evidence into the conversation about best practices and instruction, and then by extension, curriculum and other materials that enter the classroom.

Over a decade later, SAP is still doing that work. Our historical website is Achieve the Core. It's where you can find lots of information about the math standards, literacy standards, examples of what curriculum can look like. And then our newer site, Learn with SAP, really focuses on the services that we offer today, which are much more focused on supporting both the supply and the demand sides of the market to understand what the evidence is telling us about best practices as they shift, change, and improve over time.

So we do a ton of consulting and work with publishers, edtech developers, philanthropies, public benefit organizations really focused on making sure that practitioners have easy access to the best possible tools in their classrooms every day. 

[00:03:59] Alex Sarlin: That involves a wide variety of different services and tools that you offer.

There's like an academic vocabulary finder. There's, uh, you know, tools to help map the math standards, a coherence map. Tell us a little bit about some of your offerings and, and h- why the sort of edtech community might wanna really understand them. 

[00:04:16] Joy Delizo-Osborne: Well, it's so funny. I sat down with an edtech founder one day, and she's like, "You know- Look at this, this amazing blog post I found, and it shows how the math standards connect.

And I said, "You know what? That's on my website." Um, thank you so much for showing that to me and sharing it with me. That's ours. You know, I think folks-- When SAP was founded, you know, the founders really had no interest in taking sort of like personal ownership or attention from the work of the standards themselves.

So the organization really stood in the background. We were the intel inside, um, curriculum and professional learning in states and districts in terms of shifting their expectations and their procurement processes. And so we're in a lot of the things that folks already think and know they need to be focused on, right?

The coherence map, the architecture behind it, right, sits inside many of the tools that folks are using. It's been available on GitHub this whole time, right? Folks are accessing tons of our stuff all the time, and we love that. What we find doesn't necessarily come along without our help, and this is where our services come along, is really understanding, uh, it's kinda like the math standards.

You know, there's like three levels of rigor, right? There's like we focus on fluency, we focus on procedural understanding, we focus on conceptual understanding. A lot of the tools that we have out in the world, people are able to leverage for procedural understanding. What are the things my tool should be able to, to kick out?

What, what, what should they look like, right? But there's a lack of conceptual understanding of like where those expectations came from and how they might sort of impact the design of the product that you're putting together. We also are a group of former teachers, principals, instructional coaches. We understand how schools work.

We're school board members, right? Former state ed leaders, and so we understand the system also that the tool is going to live in, right? And so when we're doing advisory for edtech folks, we're able to really help them understand like that might seem like the right way to tackle fractions, right? Having, you know, 30 problems kick out various kinds.

What you haven't addressed here is the conceptual understanding of parts of a whole, right? How are you gonna get this algorithm to do that for kids before you ask them to follow a procedure 15 times and feel like, check, it's done, right? So having those conversations with folks as they're founding and building is one of, I think, our key value adds for, uh, folks in the development space for sure.

[00:06:41] Alex Sarlin: I love that. When you mentioned the sort of supply and demand that you're on both sides of the, of the coin and you, you know, the supply in this case, right? It's e- edtech vendors, it's publishers, it's everybody making curriculum, and demand is the district side, and y- and, and, and teachers and everybody actually, you know, performing education all around the country.

I, I feel like you play a very unique role. As you mentioned, you were, you were sort of an intel inside. You know, the, there was-- it was sort of behind the scenes and now, especially in this age of AI, I think Student Achievement Partners, SAP, is getting more visible, and some of the work you're doing is, yeah, is really making a difference.

T-tell us about how you've begun to address AI-generated content. We know that AI-generated lesson plans are everywhere, AI-generated assessments and question sets are everywhere. Like you just, you know, mentioned you can kick out a thousand questions now. How are you thinking about how AI-generated content is-- can stay at the high level of rigor and quality that we all want for our students?

[00:07:37] Joy Delizo-Osborne: I think it's tricky. I, I think a lot about this, like techno pragmatism sort of approach. I'm a daughter of a software engineer, and my dad was a traffic systems engineer. When I was five, he had me sitting on a Tandy, you know, with like a program he had made, uh, to help me learn my state capitals, right?

Like I have no problem with technology. My first job in the South Bronx was at a school called East Bronx Academy for the Future, and we had laptop carts and it was very exciting. So I've, I've always been very pro-tech, and I think it's one of the primary sort of accelerators, particularly for students who are historically less able to access opportunity in our country to accelerate themselves, right?

To learn on their own, get ahead, et cetera, et cetera. So I think there's so much value in it, and I think that when I was a charter principal, when I was a district employee, we are moving at a million miles an hour. Yeah. And we have a million problems to solve all at once and not enough money to solve them.

And often what that means is we're picking up the first thing that someone tells us can help us. We don't have a whole lot of time to sit and think and like test and pilot and test again and give feedback, right? That feedback loop and that consideration is really hard to find time for when you're always in a state of crisis.

And so I think the thing that we think a lot about at SAP is how to inform folks in those moments outside of the crisis moment about some of the ways that they might think about what it is they're trying to do, what problem it is they're trying to solve, and then look at all the potential solutions, not just the technological ones, right?

For the right fit for the number of resources and the number of opportunities they have at that moment. So when I think about sitting down with my local public school and they say, "You know, our curriculum is no longer going to be free, the one that we've been using for our schools." What are we gonna do?

Like, I have to cut the budget by 5 million more dollars already, right? And now there's this expense I can't cut coming in at an unexpected time. What am I gonna do? What, where do I pull the lever? There's a world in which I could say, you know, you could probably build something decent in Claude. There's a world in which you can say, "Ah, I mean, you need free.

You could probably find like close to free," right? And then there's also a world in which we say the reality of that output is unlikely to really be at the level of support your teachers need, right? So even if you kick out 300 lessons for eighth grade literacy, you give it the books, you give it the standards, there'll be lessons.

Like, where is the PL that's gonna support your teachers in executing that effectively? Where is the quality control on each of those text-dependent questions to make sure they're actually driving kids towards the right outcomes? Where is the alignment to your end of year assessment, right? There are all these pieces in there that may be available, but on a tool that's commercially available, but you wanna have that list of criteria before you go looking, right?

And then, you know, you wanna be able to match that list of criteria to the real cost of implementing with fidelity. You know, having a tool doesn't mean you're gonna use it well. So really thinking through the steps of like, if I purchase this or if I, you know, pull this into the system, if I make it available to my teachers, what are the steps I'm assuming they're gonna take to get to the point where students are getting instruction that I feel really good about?

And so I think that that whole sort of universe of questions is one that systems aren't unused to asking. They ask them about other kinds of solutions too, but I think we feel so much urgency around student outcomes that we're als- often moving even faster, you know, and the promises are big, right? These, you know, folks pull out an RCT, whatever their ESSA like tier three evidence is, and you say, "Oh, wow, okay.

That would be great. That's what I want," right? But then there isn't the level of rigor of like- What kinds of students were in this classroom or in the sample group? Right. How large is that end size anyway, right? Like, what were the supports around it? Like, there are all these questions that I think we as a, a system need to be doing better to answer for and with folks before they engage in, uh, large scale contracts.

So that's one of the things we're thinking a lot about. 

[00:11:50] Alex Sarlin: Yeah. And, and I'm hearing all of these criteria that you're naming that I think the, the edtech world really are, are working to get their head around, right? I mean, there's cost aspect, of course, there's r- there's rigor and s- and alignment to all sorts of different things, including end of year assessments and standards.

There's quality and efficacy, and you mentioned that as a tier three. But like, you know, all the edtech companies right now are figuring out exactly how to prove their efficacy because they all want it to work, but, you know, sometimes that's difficult to prove with, with RCT. And then there's implementation fidelity.

Once you've actually bought a product or worked with one, how does it actually make a difference on the ground and not just be something that is, as you say, kicking out, spitting out lots of material, but you don't know-- but it's not enough. It's not actually way designed to be implemented the right way.

You mentioned Claude as one of the, you know, go-to places for this. Obviously, Claude is a frontier model, incredibly sophisticated model. I think, you know, edtech vendors, a lot of them want to be that tool, right? I mean, they want to be the one that says, "I can do something that is actually aligned, that is actually grounded in everything you do, that includes professional learning, that has everything you need."

What would, what advice would you give edtech vendors? You know, y- you say you're on b- you know, as you say, you're on both sides of the, the table. It's like, you know, you have the schools trying to make sense in this under pressure, budget pressure, time pressure of this complex world, and then you have edtech vendors who are trying to create solutions that can actually solve those problems.

How should they think about this complex world? 

[00:13:16] Joy Delizo-Osborne: Well, I think there is, you know, a growing movement to build, right, like public purpose infrastructure out there to support developers, right? So we are one of the primary collaborators on some of the Learning Commons tools that are specifically directed at developers.

That's probably the most concrete way in which we are trying to support this piece of the industry at this time. And so we've really been supporting their work on the evaluators in literacy and really pushing, I think, the needle on sort of the reality that learning is not, it's not cut and dry . You know?

Like, one of the challenges we face is how do you know exactly where in the course of a school year's rigor a, a text might be, right? And the answer is, like, that's not really a way we can think about it . You can't say like, "This paragraph is, you know, September, and this paragraph is October," right? Like, there is something, um Sort of beautifully imprecise about, uh, the ways in which our brains work and the ways in which students learn, the ways in which we are able to apply our learning, the ways our brains sort of take what they've gi- they've been given and, and spit that back out into application, right?

And so I think there's lots and lots and lots that we are and can do as a sector to ensure design-level support for rigor, for appropriateness, for saf- you know, data safety, for engaging with kids in appropriate ways for their developmental levels, all of that. And then there just has to be A way to watch it.

Like take it out into the world and see what happens, and not in one setting or two settings, but in many, and really gather that information and start to make sort of broader sets of understandings and outputs around, like, where do these tools best fit? And so I think we're very much at work on, uh, the supply side, you know, thinking about that base level set of foundations for, like, base level quality support.

It's one of the things we say, too, we supported the publishers' criteria, right, when the Common Core first came out. So many of the Ed Reports gateways are based on some of our foundational work at SAP. And we always say, like, the Ed Reports review is the floor, not the ceiling. Right. So I think what we're, we're building right now is the floor in the AI space and the edtech space.

Like, what is the baseline level of quality you have to have to even be considered, right? But that doesn't give us the ceiling in terms of, like, how much better and more finely tuned these tools could be to actually support real practitioners and real students in real settings. And I think that's just something we're gonna have to keep working on.

Like, how do we build that more collective evidence base to understand what, which design principles work best in which settings with which limitations and which supports? 

[00:16:14] Alex Sarlin: We're in this really interesting moment now, and, you know, you talk about we have to watch how these things are implemented. There has to be-- It's not just about for any tool, whether it's a Frontier Lab, whether it's an edtech vendor, whether it's a publisher, you know, putting out the content, putting out the, the, the lesson plans, the assessments, professional learning even, and saying, "Okay, here it is.

Now it's gonna work," isn't enough, right? We have to close that loop. Yeah, it has to say, now how is it working? And, you know, when I hear you say that, there's, there's a number of different ways we've traditionally tried to do that, right? I mean, there's, there's collaborations, there's pilots, right? As you mentioned, there, there can be pilots, but they can be slow.

There can be randomized control trials or various kinds of quasi-experimental where you say, "Okay, how did this actually work? Did it move the needle? How did teachers respond to it? Did they implement it?" That, you know. There's that, which is great, but also a little slow. There's, there's usage data, right?

There's just, like, actually trying to figure out what is happening. And some edtech companies have some visibility into what's happening on the ground, but certainly not all, right? They don't, they don't know what the teacher is doing in the classroom. They don't know what is actually happening in many cases.

They know the pieces they have access to. And then, of course, as you mentioned, data, data security and privacy, there's a, I think, some, some fear about companies, whether they're Frontier Labs or edtech companies, sort of seeing what's happening in the classroom, you know, seeing too much of it. Obviously, a lot of it, it must be, you know, has to be behind all sorts of data privacy.

So, like, we have Like, I-- this is all to say, you know, I think closing that loop, having that feedback loop of actually, you know, what did, what did we create for our classroom? What sy- system of interventions and supports and professional learning, uh, tools, and then how did it actually work at feeding back into the, the originator of it is, is like the gold standard.

It's like what we all need in edtech, but it's traditionally been really hard to get. It's been slow, and, and in, in a moment of, of AI changing so fast. And, you know, we have numbers like Microsoft, uh, just put out a report on AI in education, said that 92% of teachers are using AI. So you have everybody using it, but they don't know if the quality is there, right?

They're-- we're all working on the quality floor and there's no visibility. So, like, this is a tricky moment, I think, for, for everyone. How would you address it? How can we close that loop? 

[00:18:25] Joy Delizo-Osborne: Well, I think that description was just so spot on. You know, I was thinking, um, I was on a panel recently somewhere else, and someone was asking about-- she's a first grade teacher, and she was saying, "I feel like if I'm not teaching my kids how to use AI, I'm leaving them behind."

And I'm like: Your kids use AI. My kid talks to Alexa. When we say 90-whatever percent of teachers are using AI, there's any number of ways that that could be true. Now, when I ask Google something, the first thing I'm using is AI, even if I didn't ask for that, right? It comes back with an AI answer, and it's asking me questions, right?

And so we're all using AI. It's true. Always, all the time. I mean, my microwave is probably using AI. I don't know. It's everywhere. The question is, how do we help, one, folks understand that it's not magic? Like, it came from somewhere. It can be wrong both, like, at a, a base level, right? Like, "Oh, the answer to that question is incorrect."

But it could also be wrong in terms of, like, the way it is making sense of the inputs it received. I just saw something on LinkedIn. A woman was talking about Gemini, and how it ended up admitting that it had mischaracterized work on her resume it had asked her to write because she was a woman, and she had all these face-- Uh, she had all these screenshots of actual, like, self-reflection on Gemini's part.

Like, "I did use different verbs for you than I would've used if you, if your name was Jeff as opposed to Jennifer," right? Like, there are ways in which we only really have biased data sets because we're biased people. The algorithm can't outthink the data set it has been given and tuned to, to leverage.

And so I think, like, there are lots of ways in which we can think about the different challenges that AI presents and the opportunities, which are vast many. And then I think we think, yes, how do we start to sort of understand the very complex set of variables that are working together in a school system to get to the outcome that we want to see?

Because rarely is one tool the only thing, right, that is an input in a student's learning. So even when you show a certain number of years of, you know, change in student outcomes, you have to understand what else is happening in that system to get there. Are they also using a high-quality instructional material on top of that, right?

Are they having, you know, tutors? Whatever the other interventions are in a system, I think it's really important for us to start to understand what they are and how they interact with one another. Some of our friends at the Gates Foundation are, you know, really beginning to think about things in a stack sort of way.

Like, what's the full stack that works best together to enter a system in a particular state? And so I think that as edtech providers continue to innovate and think about their individual tools, they also wanna think about the relationship of those tools to the other tools that are likely to be in a space and really start to think about a full stack or, like, a, a technological environment, right, that involves both Sort of typical computer-based tech, but also the technology of human beings.

What are the other supports that are happening in a system, or that you need to tell a system to use our tool really well? What we see are these three supports around it really supporting student growth. And I think getting to that kind of clarity is important. It's one of the things that SAP is really thinking a lot about.

How might we support the sector in starting to gather more information about the human practices that are happening in and around classrooms to support ed tech implementation, but also core curricular implementation, and how can we make those practices more visible to systems before they pick up the tool, before they procure, before they roll out?

What are the s- the sort of environmental enabling conditions you need to have in place to make sure that your teachers are able to actually use this tool to get to student outcomes? Time on task doesn't tell us everything, right? And so how do we get to a place where kids are ready to spend 20 minutes with Amira and actually see the outcome that you want them to see on the other side?

I think that's a more complex question than, you know, a 20-minute intervention block on the calendar or in the schedule. Trying to answer that question more holistically, I think, as a sector will really serve us even as the tools improve and change quickly. We can continue to improve product quality, but if we don't support systems in being able to use them, right, the quality kind of just sits there, right?

It doesn't do anything. So I think that's part of the s- the solution we're trying to build. 

[00:23:07] Alex Sarlin: And we've, we've seen a lot of studies recently about how the dosage is just tur- comes out much lower than, than recommended and wanted for ed tech tools. They say, "Okay, if you use it this much per week, you should see results."

And then they come back and say, "No, they used it a fraction of it that week." I mean, that's been a c- a common issue in ed tech for a long time, but now that loop is, is getting very clear. Y- I mean, the s- such good points throughout, you know, everything you're saying there. Just s- the combination of, you know, technology trying to sort of solve things in this very direct way, and then this very complex educational environment with many different inputs.

P- students are doing things at home. Teachers are doing all sorts of different things based on what classroom you're in. There's professional learning. There's all of the, there's high quality curriculum or what quality curriculum is being used. So many different factors. You mentioned earlier that you're a techno pragmatist, and I, that word jumped out to me because I think in this context, we're looking at everybody's looking for, for a silver bullet that's gonna sort of solve it, that's gonna make a difference, especially ed tech companies, that's exactly what they, they wanna be that silver bullet.

And yet when you get into the classroom, there's so many factors that complicate things. When you say you're a techno pragmatist in education, what does, what does that mean in terms of how to combine the sort of ideal, you know, vision of what, what entrepreneurs in ed tech want with the practicality of what the classroom looks like?

[00:24:27] Joy Delizo-Osborne: I mean, I would love for folks to spend more time in classrooms. We don't all work in schools, but I mean, I have kids. I, I know more about my daughter's learning experiences from spending 10 minutes in the hallway just listening than I do from my 20 years in education. I think that there is a lot to learn in just being present and understanding the reality of the challenges that folks are facing.

I mean, as a charter principal in Oakland, where we were a one-to-one school, any, any and every, you know, technological tool available to me and my students, and, you know, some days there's brown water coming out of the ceiling, and then the Chromebooks get rained on, and then you're starting over, right?

Sometimes a kid takes a Chromebook home and doesn't bring it back. Sometimes ... Like, there are a million, a million things going on in buildings all the time, and understanding how hard dosage, uh, for one tool can get, can be to get to, much less like three different ones, right? Three different subject matters, three different foci.

The challenge in a day starts so far beyond just, like, using tools. So it's like, how do we get them to school? How do we get them fed? How do we get them to sit down? How do we get them to be kind to each other? How do we get them to focus, right? There are all these questions that a teacher is answering, and I think the more that systems are able to sort of have a, a tool or a product set come and say, like, "This is what happens when things break down, and this is the backstop we have," right?

"This is the bare minimum you can use this for. This is the way that it can be worked with another tool in your product suite to help sort of fill that gap. This is how it fits into your core curriculum," right? And so one of the tools that we really appreciate and one of the efforts we are supporting as a collaborator, AI for Equity, has a landing page, hqim.ai.

And I think the real push there on Erin's team is like, why aren't we using more of what schools are already using, right? Like, why does your tool need to be a full standalone? Mm-hmm. Can it speak to, work off of, build on, enhance something that its teacher is already going to have in hand? And by that sort of token, be a natural extension of what they're already trying to do as opposed to an add-on, right?

Are there ways to think about core curriculum as a starting point, regardless of the kind of tool that you're trying to build, so that you're able to build around what already has to happen and, and not feel like a burden? I think there are also ways in which folks can think about out of school time, extracurricular time, summertime.

My daughter's in a summer camp right now where they use Amira, right? And I was like, "Oh, that's so interesting." She's like, "Yeah, you know, I do my reading practice." And like, what a, an amazing way to take, uh, pressure off of a summer school teacher and have a tool that can meet kids individually where they are and push them forward.

Like, that's a great use, right? So thinking about not necessarily putting all of the pressure on the core instructional day or outside of the core instructional product, and thinking about ways to enhance what's already happening and what's mandated, honestly, for, for systems, I think is, is really key. I don't know.

It's a saturated marketplace. Yeah. It's a busy one. There are lots of options, and I think the more we can think about Combining and lessening pressure for individual practitioners, the better. I mean- Yeah ... this is where Magic School, you know, and some of these other solutions are so loved by teachers. I have lots of teachers in my life who use AI, not to supplant their own lesson planning, but to enhance it.

I've been doing this for 20 years. I know exactly what I'm gonna say and do. What I need is five practice sheets Sure. For that practice teacher, it's a perfect solution. It keeps them from having to do a bunch of sort of manual labor because they already know how they're getting to conceptual understanding.

So the fact that what comes out of AI is just procedural doesn't matter. When we start to think about an AI pro- program as the full teaching experience or learning experience for a student, I think that's where things start to break, right? Just how well is it tuned to do all of the things that a teacher with lots of experience, tenure, and knowledge of just, like, kids would do, you know, naturally of their own accord, so.

[00:29:07] Alex Sarlin: Yeah. No, it's an amazing point. I, I-- What I picture when I hear you talking about it in this way is like, is like a, the handshake between where the technology sort of, it leaps off and where the, the implementation and the, the teacher, the teacher's previous work and their behavior sort of comes in. And we have seen a little bit of interesting movement in that space, I'm sure you know about.

You know, like, you have, like, CourseMojo built directly on top of HQIM materials, or CoTeacher, or, um, you know, both Brisk and Magic School are doing a lot of work to sort of ground what they're doing in existing curricula so that they say, "Okay, if you're already working with this core curriculum and these supplementals, everything we do is gonna come out of that, that co- corpus of material."

I feel like there's been some really interesting movement, especially just over the last, like, year or so, in edtech tools doing exactly what you're recommending, saying, you know, we don't have to solve everything from scratch and just the classic, you know, "Education is broken, and we're gonna come fix it."

It's, "Hey, there's all of this stuff already happening in the classroom. There's-- If we can sort of tap into what, what is happening but is not, but, but is not perfect yet, you know, is not, and enhance it and make a difference, you can really move the needle." And as you say, you can see adoption much faster because you can see-- because you have people who are already halfway there.

You mentioned Aaron Cuen and AI for Equity, a really amazing organization. You're, you've also done some interesting reports in a collaboration with AI for Education about how districts can make sense of AI tools and evaluate what they're doing in the classroom. Tell us a little bit about that work. I think it's really relevant to this conversation.

[00:30:35] Joy Delizo-Osborne: Absolutely. Yeah. So our-- We have two reports. One is on sort of ensuring rigor in mathematics. And I mean, in both of those things, we really think about supporting a district and making sense of what are the gaps in their current instructional practices and routines across their s- their system, right?

What, what's happening in schools that you need to change? How do you know? Like, actually go find out, you know. Where, where are your teachers really struggling to make connections and support students? And then how do you evaluate the tools you're potentially bringing in for the value that they might add in those areas?

So when we think about the, the value and beauty of differentiation of texts for students in a classroom, we also wanna think about the fact that all students should be expected to be doing grade level thinking Right. You have to maintain a productive struggle. So what does that look like in the tool that you're bringing in?

How might, in a demo, you actually see right, what it looks like to be different students at different levels, but not have AI step in and take away the struggle for the lower level student, right? What can someone show you that demonstrates that the tool does not, in fact, take away the opportunity for cognitive thinking and application from students?

I think the other things we think a lot about are, again, in mathematics, that difference between fluency, procedural understanding, and conceptual understanding. In what ways is the tool bringing real-life experiences to the front for students to conceptually understand the value and the point ultimately of learning the thing that's in front of you?

I mean, when I was a teacher in a while. But you know, the big thing then was, was Khan Academy. That was, like, so exciting. I'm like, "Oh, I can just put them on, and they'll just practice, and it's great." And Khan is great. I was an SAT tutor too, right? Like, the amount of resources that are available on Khan Academy are amazing, and it's not great to stand alone, right?

My daughter uses Khan Academy Kids often, right? But I wanna be there too, asking additional questions, making sure she's actually understanding. You know, when they're talking about three hats, you know, I'm like, "Well, go get us a hat. Can you go in the closet? I wanna see your hats. How many hats do you have?"

Right? Like, how do we take it off the screen and into real life and have things really feel like they're worthwhile, it's work worth doing? That's where we start to lose kids in terms of relevance down the line, and that's part of what those guides are really focused on, is how do we bring in, uh, AI, not just to sort of lighten the teacher's load, right?

But also to actually build student engagement and their sense of relevance and value in the education they're receiving. So both of those reports offer really great nuggets. There's a lot to think about, and so I think that's one of the things we really appreciate about folks like AI for Equity who are building, like, really durable frameworks for thinking about this stuff.

Picking up the framework and looking at the tool in relationship to it can be really helpful, and then also thinking about how to build just, like, AI literacy. We obviously talk about literacy as in, like, reading, right, and writing, and I think there is also this piece of understanding when leveraging an AI-based tool is actually useful.

And I think adults are learning this too, right? I was talking to someone else. They were talking about an independent school where their student attends and how they have a red, yellow, green framework for assignments where, you know, green is, "Go ahead. Sure. Use whatever tool makes this more doable. I believe you've already addressed the conceptual understanding, so what I want you to do is to be able to use AI to m-make the process faster."

Right? And then all the way down to red, which is like, "You don't have an understanding of this concept yet." Your brain should be doing all of this work . Right. Like, this is not a moment to go to Claude, right? Like, you stay with-- stay inside yourself and, and actually, like, confront where it is you understand what's going on, where you don't, and come back and report to me where you need assistance so I can push your brain's thinking before you move into the sort of technologically assisted moment in this process.

And so I think that's actually where we're gonna end up as a sector, right? Is starting to be clearer about where kids need to engage on their own with concepts, and then as they master conceptually, procedurally, can explain, can engage, sort of move up the ladder to technological support, because it really is about adding efficiency.

It really is about sort of saving time, building knowledge about new ways to apply the same concept over and over and over again. I am hopeful that we'll eventually get to a place where we don't necessarily expect AI to be in every experience or tech to be in every experience that students have in a, in a school day.

But we're really clear about, one, what's building their literacy in terms of technological advancement, and then two, where these tools can just make the learning process easier, faster, more efficient. One of the things I do on the side, I, I'm a college, uh, essay coach. It's like I can't stop. I, I love working with kids on their writing, and it's like an obsession.

And a lot of people are like, "You have a really big job." I'm like, "I do." And I, I work with kids on essays, and I think it's one of the most amazing things 'cause it's like you can't farm it out. Like you just... You can't. I mean, you can. It just won't go well, right? Like, ultimately, writing about yourself is something only you should be doing.

And so it's a really beautiful place to like have kids be like And like the way AI can be useful here is as you're talking to me about all the things you've learned in high school, the transcriber is gonna go, and it's gonna summarize some of those themes. And at the end of this conversation, we're gonna have like four or five things that we're gonna go and mine deeply for things to write about, right?

But it's not gonna write it for you, right? It can listen and tell you what you thought. It can tell you what you learned at summer camp. It can tell you, you know, all these things from what you've actually said, but then you have to go and put the narrative together. And so I think there is this piece for all of us, whether it's the teacher, whether it's the district leader, whether it's the student, w- what are we giving and what are we getting back?

How can we be clear and clearer about what we need to keep as work for ourselves in order to get to the outcomes that we want to see? I think this is a super exciting time. I think there's so much to learn. I just wanna make sure we're not learning it on the backs of the kids who are least privileged.

You know, I think independent schools take their time because they can, you know, 'cause their kids come to school knowing lots and having lots of resources, and so I think they, they ha-- I went to a really great independent school, and they were thinking about a lot of really interesting things in terms of AI, but none of it is to replace the thinking or the struggle for students, right?

That focus on struggle is core to their mission. You're gonna have a hard time here so that when you get to Harvard, it feels like a day in the park, right? The, the struggle is a key piece of the promise they're making to their constituents, and so they make sure it's still there. I think that is not the same, right, in a much more sort of diverse set of public schools across the nation.

Like we are, we're just trying to get it done 'cause with less money than we need, with fewer resources than we should have, and so we're looking for shortcuts a lot more often, and I think That's where things get tricky, you know, and we're gonna need to be careful. I 

[00:38:28] Alex Sarlin: love hearing you talk about this, and I feel like that you're, you're, you're di- diving into some of the absolutely, like, core tensions I think that exist in the, in the ed tech, in the ed tech and education space between different types of schools, between different types of learners.

One that I frankly have really been wrestling with recently, and I think you, you just spoke to it, so I'd love to double-click on it, is this idea of when does the AI get involved? 'Cause we, we agree. I mean, your, your, the reports that, that, um, that SAP has put out, they talk about, you know, productive versus counterproductive struggle, and how to sort of keep the thinking on the student while they're learning.

It makes, that makes so much sense. In my instructional design hat, I say, "Of course, definitely." At the same time, we all know that the entire value of AI is that it's so fast, it's so responsive, it's so iterative, it's, it, it does things instantly. And I think, you know, you also see other reports or other approaches saying, "Oh, start with AI, it can do this whole thing and then analyze it."

I, I remember talking to a, uh, a coding entrepreneur a, a couple of years ago who said, "Look, we can teach coding in a completely different way. Instead of starting from 'Hello, world' and building up, you can start by saying, 'I want to build this amazing game,' and it can build it for you, and then you can deconstruct it and actually learn."

So that's the mode of AI first, right? So there's this AI first mode where you build engagement and excitement and jump over all of the, the painful early learning steps. And then there's the-- But, but, but I, but at, at the same time, we know from, from pedagogy that that could take away all the productive struggle, all the work, all the conceptual understanding.

I, I really wrestle with this, and I think that the types of frameworks that you're recommending that we sort of need to come up with, it's, you know, at what age, what types of-- what are you learning? If you're learning a core concept, do you want to learn that on your own or from a human teacher, and then only g- uh, you sort of earn the right to use AI by understanding the concept?

That's a really interesting model. Versus, hey, the AI can give you 20 different metaphors and a visual and a video and make it so interesting to learn that concept on day one that you're excited about it. That's a totally different-- I mean, it's almost exactly the opposite. I'm curious how you reconcile these because frankly, I am, I, after interviewing many people from lots of different sides of this, I'm really wrestling with it.

[00:40:36] Joy Delizo-Osborne: No, I think it is-- I mean, it's the question of this moment. Like, I don't have an answer. I do have a set of observations at times. Like, you know, as an adult, I've had a bajillion experiences already . Right? So having AI jump ahead and then to go back and deconstruct doesn't undo the fact that I had to learn a bunch of things myself first, right?

If I were my child, who's three, almost four, right, and, and, and her brain never has the opportunity to work Forwards as opposed to backwards, right? That is a fundamentally different set of expectations, starting much younger at a much earlier stage of brain development. Like what I don't need or what she doesn't need.

And so I think that to your point, right, it-- there is something to, to be done and, and, and I maybe we should do it. I'm, I'm like, huh. But thinking a lot about... I, I, I was at a presentation on, on the science of learning from Jim Healy a few months ago, and, you know, our brains learn by struggling, like that's how they develop.

Without challenge, they do not, right? They atrophy in other ways. So like there is no world, I don't think, in which there isn't value in, in starting from scratch, you know? And, you know, we think about attention spans of our kids, our-- their ability to sit with discomfort and challenge. I grew up in a really religious home, and one of the things I worry about for my kids is like, I learned how to pay attention from sitting through sermons on Sundays.

I learned how to sit through lectures in college, in church when I was, you know, seven, right? And so, and it's not because, like, I needed to know exactly what was happening in the sermon. It wasn't about the information, it was about the experience of sitting quietly and being, you know, knowing that I would be accountable at the end for what happened, right?

And so it's like, how do we know what the kinds of experiences are that we as humans need? I mean, we do know. There's lots of s- there's lots of research on the science of learning out there. It's actually not a mystery. So how do we actually pay attention to those, even as AI promises faster, easier, better- Right

more efficient, right? How do we say those things are great, right? Like I can drive a car, I can fly a plane. It doesn't, uh, invalidate the value I get from taking a walk, right? Like all of those things can be valuable in different ways. I wouldn't walk from here to Shanghai, fair. But I wouldn't take a plane from here to the grocery store a mile and a half away either, right?

So like how do we decide what is the right tool for the situation we're in? I think we're so excited about the possibility right now that naturally, humanly, we wanna use it everywhere. And it's like, well, or we could say, you know, this tool is for this situation and this tool is for that situation, and we actually have a, you know, age-old tool, like a stick and a hammer, you know, the-- and they're the right thing for this tool over, or this problem over here, right?

And so I think in five more years we'll have a better sense, you know, once some of the shine has worn off of, you know, where we feel this set of tools is most useful. I think we have to go through a lot of trial and error to get there. Again, I'm just hoping We can do that in a way that, that doesn't leave a bunch of kids without the cognitive development that they deserve and need to get from school, because that's what we've promised our families.

That's the promise of public education in this democracy, right? Is that our schools are delivering folks who are able to think critically and to make sense of the world around them in ways that would support this coun- this country in being successful in the long run. And so I think that there are just a lot of questions, particularly in my mind, younger kids.

I know lots of folks are thinking a lot about the ways in which AI and tech can support adults In those spaces and being better trained, m- more knowledgeable about best practice, right? Places where salary levels are low, value in higher education is probably not there for the person, right? There is no ROI in, in some cases.

And so how some of these learning opportunities can be accelerated by tech as opposed to putting the tech in front of the small child, right? Like making the product that's getting better is actually the one that is refining the teacher knowledge and practice, not necessarily the student experience.

And so I think that there are lot-- there's lots of promise there. Co-teach, similar, you know, idea, right? And I think that CourseMojo, I love that team. I think they're thinking really intentionally about the ways in which Mojo supports students, but then also is supporting the teacher in doing their work better, right?

Yeah. So it's not just a student-facing tool, it's also a reporting and support tool for the teachers so that when they are engaging directly with kids, they're doing it in the most efficient and thoughtful way possible. And so I think there's these emerging models that are super exciting. And I, I don't know.

I'm excited to see what happens. And, you know, we try to be in buildings as much as possible to also see all of the other stuff, right? The full stack, right? Of interactions across the day, and really think about where the plane that is AI would be most useful. Yes. For sure. Yeah. 

[00:46:09] Alex Sarlin: And I think you're, you're really well-positioned, sort of uniquely positioned because you've been-- SAP has been playing this role in helping, in sort of bridging practice and, I don't, I won't say theory, but in bridging, you know, tooling and practice, standards and practice, you know.

The vision of what it could be with how it's going to actually be implemented in the classroom, and it feels like AI is at this moment that desperately needs some thinking around that. People are setting these age limits, right? They're saying nobody under X years old should be using AI. Personally, that doesn't quite resonate with me.

I, I get why people c- get there though, because it, it's a, it's a proxy for exactly what you're saying. If you're younger than a certain age, your conceptual understanding may-- or you're, you're not ready to sit still and, and listen. You're, you know, all these things are sort of, there, there's a proxy there.

But I think we can get more nuance on that about, you know, what type of learning experience, what type of learner, when is AI, you know, useful, and when is it actually a crutch or a, a, a detriment to learning? Because we know it is. 

[00:47:06] Joy Delizo-Osborne: And we can teach kids to be-- I mean, one of the things I was talking to someone about too is like I, I was one of those kids that went across town for school my whole life, you know, tested into various things.

And I think the thing that I, I learned the most, like the most important distinction that I experienced, was some people are taught how to be agents of their own experiences and futures, and other people are taught how to follow instructions. And The first thing I, that comes to mind when I think about my very young child and AI is like how does she act as an agent?

How does she see AI as something to be directed to get to the outcome that she wants as opposed to the teacher that knows more than she does that is there to give her instructions, right? And so I think that there is also just like a stance that we can have, a, a, a positionality that we can have, an expectation that we can have, that these tools position the learner as the agent and then our job is to teach the agent h-how to have agency, right?

Right. And so, you know, it's like the whole point of AI building something and you deconstructing it is saying like, "I'm looking for what you did here, and I'm actually checking to make sure it works the way I wanted it to." But in order for that to be true, I have to know very clearly what I wanted, right?

Right. And that's the piece that I get worried about when people are like, "Oh, like have AI do it for..." Like, but, but if I don't know what I was looking for, how do I check, right, for quality, for, for insistence on, on the outcomes that I want to see? If I haven't actually fully built out that picture 'cause I don't have the experience or haven't taken the time or whatever it is, right?

Ultimately, I'm evaluating with a very thin set of guidance and expectation that ultimately leaves us with a lot of like mid kinda stuff. You know, it's like mediocre. And was that then actually any better than just doing it slower to start with? Like a bunch of mediocre outputs are not actually better than a few really high quality ones that took a little longer to get to, right?

And so I think that agency and that agent position I think is one that we as adults assume everyone will have because we-- this is a new toy that showed up for us to make our lives easier, right? But when you're an AI native, right, your positioning and your sense of what the tool is and how it came to be and what it has to offer are gonna be very different, and I just don't think we know enough about how Kids' minds will be shaped by those experiences just yet, so 

[00:49:50] Alex Sarlin: 100%.

Really insightful. I think that I totally agree. In this conversation, we've sort of hit many of the major challenges of this age, both with regards to AI and just with regards to implementation, right? How, how do we actually get to the, the results we want given all the visions, all the competing visions, all the different factors that play into any education.

This has been a really interesting conversation. I could go another hour for sure, but- I 

[00:50:13] Joy Delizo-Osborne: know 

[00:50:14] Alex Sarlin: We are running 

[00:50:15] Joy Delizo-Osborne: out of time. We're at time. I know. 

[00:50:17] Alex Sarlin: So much fun. Joy Osborne is president and CEO of SAP, Student Achievement Partners. This has been a blast, and wow, I'm leaving with a lot to think about and I'm sure our listeners are as well.

Thank you so much for being here with us on Edtech Insiders. 

[00:50:30] Joy Delizo-Osborne: Thanks, Alex. Thanks for having me. Have a great rest of the day. 

[00:50:33] Alex Sarlin: You too. Thanks for listening to this episode of Edtech Insiders. If you like the podcast, remember to rate it and share it with others in the edtech community. For those who want even more Edtech Insider, subscribe to the free Edtech Insiders newsletter on Substack.

This season of Edtech Insiders is brought to you by Starbridge. Every year, K-12 districts and higher ed institutions spend over half a trillion dollars, but most sales teams miss the signals. Starbridge tracks early signs like board minutes, budget drafts, and strategic plans, and then helps you turn them into personalized outreach fast.

Win the deal before it hits the RFP stage. That's how top edtech teams stay ahead.