Week in Edtech 6/17/26: OpenAI, Google & Anthropic Expand AI Workforce Training, UK Weighs Social Media Ban, DonorsChoose AI Requests Surge, and More! Feat. Tom ap Simon of Pearson & France Hoang of BoodleBox

Join hosts Alex Sarlin and special guest co-host Gemma Lenowitz Picot of Samvid Ventures as they break down the latest developments in AI, workforce development, edtech adoption, and education policy.

✨ Episode Highlights:
[00:03:26] OpenAI, Google, and Anthropic expand AI workforce training initiatives
[00:10:30] Data center investments create new opportunities for workforce development
[00:15:59] Anthropic launches Claude Core to help nonprofits adopt AI
[00:21:00] AI safety concerns raise questions about education-specific AI models
[00:27:23] Khanmigo tops the new EdTech AI Visibility Index
[00:33:40] Teacher requests for AI tools on DonorsChoose have tripled
[00:38:26] AI adoption may accelerate growth in microschools and school choice
[00:41:22] UK proposes banning social media access for children under 16
[00:42:37] UK expands AI tutoring access for low-income students
[00:50:18] Satya Nadella introduces the concept of AI-powered organizational learning loops

Plus, special guests:
[00:54:02] Tom ap Simon, President of Higher Education and Virtual Learning at Pearson, on AI readiness, career pathways, and workforce outcomes
[01:16:03] France Hoang, Founder and CEO of BoodleBox, on AI collaboration, financial literacy, and preparing learners for an AI-powered future 

😎 Stay updated with Edtech Insiders! 

Follow us on our podcast, newsletter & LinkedIn here.

🎉 Presenting Sponsor/s:

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.

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.

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.

Transcript

[00:00:00] Alex Sarlin: 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. 

[00:00:31] Gemma Lenowitz Picot: These models need workers to build data centers and HVAC and plumb them so that their financial projections will satisfy investors, and that means there is a lot of opportunity for workforce training that can equip individuals from low-income backgrounds into these technical roles.

The question is whether they are highly professional roles and what the career navigation options look like. And I think we're seeing nonprofits like Per Scholas enter this space with employer relationships and really understanding the job seeker and having experience designing curriculum for somebody who has maybe a high school diploma or an equivalency and not necessarily a college degree.

[00:01:13] 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. 

[00:01:28] 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.

[00:01:52] Alex Sarlin: Welcome to Week in EdTech. We are here the week of June 15th, mid-June, height of summer. School is basically out, definitely for higher ed, uh, certainly getting there for, for elementary school all over the country, and we are still seeing so many headlines and so much happening in the edtech field, especially around work.

We have an incredibly special guest host today, one of my favorite people in edtech, and a, a core member of the Edtech Insiders community, Gemma Picot from Samvid Ventures. Gemma, welcome to the pod. 

[00:02:23] Gemma Lenowitz Picot: Thanks, Alex. It's incredible to be here. I'm a longtime fan, and I think I sent you a Slack message in 2021 saying, "You're really onto something.

You should get a, a real microphone." 

[00:02:33] Alex Sarlin: I think so. 

[00:02:34] Gemma Lenowitz Picot: It's an honor to five years later be joining you. 

[00:02:37] Alex Sarlin: That's so nice. You-- I think you did. And I did, and yet I don't use it every time. People may notice sometimes my mic is a lot better than others. Well, it's great to have you here. Before we jump into the headlines for today, give, give our listeners a little bit of an idea of what Samvid Ventures does.

It's really interesting. It's sort of at the intersection of economic mobility and education. Give us the overview 'cause it's a really cool organization. 

[00:02:59] Gemma Lenowitz Picot: Thanks, Alex. Yes. Samvid Ventures is the impact investing arm of Samvid Philanthropies. So we use philanthropic capital to back early-stage founders that are building tech-enabled solutions to drive economic opportunity for low and moderate income Americans.

So we look at workforce development, edtech, wealth building tools, benefits access, and we evaluate all of those based on whether they produce measurable improvements for low income Americans. 

[00:03:26] Alex Sarlin: Incredibly important work, incredibly relevant work to this moment, and especially at this time of enormous inequality and a, a huge admiration for what you're doing there.

It's been really fun to watch all the, all the investments and all the work that Samvid has been doing. So let's jump into the headlines. Gemma, you, you are so deep in the workforce training space, and you understand this so well.

One of the things that keeps coming up now, and it's really interesting, is that frontier labs are starting to offer more and more programs, often for low income Americans, to try to sort of upskill them in various ways. And just this week we saw OpenAI announce a whole set of free courses for their OpenAI academy.

We saw Google invest over a billion dollars in skill training for Alabama, where they're building a data center, and we saw Anthropic create this Claude Core, which is meant to teach nonprofits how to use AI more effectively. So all of the frontier labs are sort of thinking about what to do in this AI era and how they can, how they can sort of help people who may otherwise be left behind.

I'm so curious how you think about this. 

[00:04:33] Gemma Lenowitz Picot: Yes, and it's all in the backdrop of racing toward IPOs for OpenAI and Anthropic, so it's hard to distinguish what might be real from what might be marketing. But let's sort of keep a, a positive view as, as we can. I think there are a lot of opportunities for individuals to access digital learning through these providers, and of course, now they're thinking about their integration into a number of enterprise contexts.

So I think OpenAI now launching courses on Academy is really a signal of their enterprise pivot, this new upskilling playbook. So they're trying to position as the default workplace learning infrastructure and really orienting more aggressively toward these high productivity use cases For economic mobility, I think the key question is often who gets access.

So if these courses stay in large enterprise channels, they reinforce existing advantages rather than distributing them. But these courses are designed to move employees from very basic prompting knowledge to running fully agent-assisted workflows. And we're hearing this from partners like Walmart as well, that are thinking about training even their frontline sales store associates using AI tools.

And so I think the labs themselves need perpetually increasing growth in token usage. OpenAI is now betting that that enterprise will help them expand beyond a consumer use case. But I think we're also gonna see increasing scrutiny from enterprises about the value proposition, and the training really has to work.

[00:06:05] Alex Sarlin: There's been a really interesting split, I think. You know, when you just sort of see people's reactions to the increased usage of the... you know, al- almost every company, I won't say almost every, but huge numbers of companies have implemented one or more of the frontier labs using Claude institutionally or using OpenAI and ChatGPT or using Gemini for institutions.

They-- people are really adopting these very quickly or, or other tools as well. We saw SpaceX buy Cursor j-just the other day. That's, uh, you know, Cursor is one of a whole series of, of coding tools that people are using en masse. So people are buying this software everywhere. We're subscribing to the SaaS platforms.

But some companies are seeing a lot more productivity usage, a lot more benefits than others. And I think to your point, some companies are probably starting to get a little skeptical. They're saying, "Okay, it sounds like we can do amazing things, but none of my employees are actually building agents. None of my employees are actually creating workflows.

Nobody can prove yet that they're getting real, real benefits." And I mean, some companies are doing more than others. We talked to Anthony Salcito from Coursera. He mentioned that companies like Moderna and Siemens were really carefully measuring the productivity increases from AI. But I think this is so new, I'm sure the majority of companies don't have that infrastructure in place.

So to your point, they increasingly need to be able to use the AI to its capacity to actually see productivity growth. It's not just having access to it, it's knowing how to do things with it that are really sort of above and beyond, and the training has to be there. 

[00:07:31] Gemma Lenowitz Picot: Absolutely. And this is where token maxing is no longer part of the conversation and, and is more focused on ROI.

But a few years ago, we were skeptical that ChatGPT wrappers were going to add value and that there was no moat from a defensibility perspective for companies that were building a wrapper. I think what we're seeing now is that user experience does matter, and these LLM providers have realized that just putting a person in front of an empty box does not yield a great outcome.

And so they're figuring out upskilling and more scaffolded and differentiated courses based on roles so that people can understand their use cases. But I don't think we've moved beyond the sort of publicly facing prompt library in a significant way Through the frontier labs. I think actually there are a lot of organizations providing more specific training on top of the sort of vertical needs that employees are bringing to their work.

[00:08:22] Alex Sarlin: Totally. I mean, I th- I've seen a little bit of an interesting UI improvements around sort of building workflows and agents. Like, I think it's still not easy, but if you look at some of the workflow tools, they're starting to become more drag and drop. They're starting to seem a little bit more like traditional, sort of easy-to-use UI, where you're like, "Oh, you need to do a Google search?

Put that here. You need to get data out of here? Just, like, connect it up with a little arrow." Like, they're starting to sort of build it in a way that looks more visual, that looks more simple, rather than it all being through text, right? Where you're just like, "Hey, I need to do these 10 steps, and I'm gonna describe each of them, and then you have to give you access to each of the tool."

Like, it's starting to look a little more like that, and I think there's gonna be actually a race. I think they're gonna realize that if agentic use is the... And multi-agent and sort of everybody has their own team of agents, if that's really where we want this to go, then they need to make that very easy to create your team.

And it is not yet easy, but it's hopefully getting there. The combination of training and UI improvements I think could be really powerful. Your point about wrappers is interesting. I wanna pick your brain about one other part of this, though, because, you know, we mentioned in last week's Work-- Week in EdTech that Meta had just announced this workforce academy where they're building data centers in a variety of places, at Baton Rouge, in Indianapolis, and a couple of other places, and then they're creating these sort of whole skill training initiatives around it and saying, "Oh, and as part of the data center, you know, initiative, you're also-- we're also gonna train all these people and create jobs, and there'll be electricians and welders and all these people."

And then this week we see Google saying, "Oh, in Alabama, we're putting a, a billion dollars, and it includes h- half a million dollars for STEM kits for schools, and it includes 130,000 Alabamians training for the..." And you're like, you're seeing these sort of package deals where big tech companies and frontier labs are saying, "How do we enter a community as not a negative, not like, you know, the bad guy in, uh, in, in It's a Wonderful Life?

Like, we-- how do we enter and make sure that we're being seen as, as a positive influence?" And I'm so curious because this is all about workforce training. Like, do you think this is legitimate? Do you think these are really gonna create skills for what were low-income employees, or do you think this is sort of whitewashing and PR or some combination?

[00:10:30] Gemma Lenowitz Picot: For now, I think there's opportunity because there's so much infrastructure build. The infrastructure CapEx is really propping up the American economy. So these models need workers to build data centers and, you know, HVAC and plumb them so that their financial projections will satisfy investors. And that means there is a lot of opportunity for workforce training that can equip individuals from low-income backgrounds into these technical roles.

The question is whether they are highly professional roles and what the career navigation options look like. And I think we're seeing nonprofits like Per Scholas enter this space with employer relationships and really understanding the job seeker and having experience designing curriculum for somebody who has maybe a high school diploma or an equivalency and not necessarily a college degree.

And I think it is the providers that have the experience with those learners that will succeed in building, you know, economically mobile pathways. And they are delivering more of short form. These are 16-week programs, often combining some kind of technical instruction with on-the-job training. They might look like apprenticeships.

And so I think that will certainly create jobs. Whether those jobs are sustained past this moment of data center build-out, maybe we'll be talking about space data center build-out one day. But we should be mindful that upskilling into any role is temporary to adjust for the labor market. And so we should give any worker more opportunities to navigate the durability of their career against other options, and ideally will beget more short form training that can then upskill broader populations.

[00:12:08] Alex Sarlin: Yeah, that's a great point. Uh, how do you see the relationship between, you know, this sort of regionalized training that you're seeing with Google and Meta in these cases, and then the sort of open training like OpenAI Academy or Anthropic puts a lot of courses online. Google has a huge suite of education courses.

They're putting them out all the time. They do them through lots of platforms. You know, so they put out all of this sort of accessible free education, but in some ways that's a real, like, pull model. Individuals have to sort of find that and see the value and stick to it and get those credentials 'cause they feel like there's a value to it and make sure that it's relevant to their career versus this sort of push model where you're like, "We're going into Alabama and we're gonna train 130,000 people, and if you're in Alabama, if you're in Jackson County, Alabama, we're gonna come really make this, like, right in front of you, and you can actually come in in person for 16 weeks and learn something."

It's like it's two very different models of upskilling, and I'm curious how you see them playing together. Do you think the regional push in is gonna be much more effective, or are we gonna get to a place where everybody needs a Google certificate to get a job, so the online Google credentials are incredibly popular?

We've already seen, you know, mil- uh, over a million people take them on, on Coursera, for example. Well, how do you see these playing together? 

[00:13:20] Gemma Lenowitz Picot: I see this as kind of a blue collar, white collar split, and it makes me think about Molly Kinder at Brookings, who's done some brilliant economic analysis around AI and workforce.

And she talks about how the jobs susceptible to automation are the opposite of the jobs that we required as essential work in the COVID era, and how if you could do your job with a laptop in a closet, that now you should consider that AI will be able to do that. And so I think I consider the upskilling of employees in a white collar context to look a lot more like training models to be able to take that capability forward.

Of course, we'll talk about where human in the loop plays a role But I think these blue collar jobs feel durable. I don't know that every person trained to build data centers will then follow on to build all the necessary housing that we need in this country as well. But I think the idea of infrastructure and onshoring more manufacturing will make those jobs more durable.

But I think the training looks quite different. And I actually am not sure how AI forward the blue collar job training is at the moment. I think it's probably integrated, but not necessarily at the, you know, teaching folks how to use agents level. 

[00:14:31] Alex Sarlin: Right. That is a really excellent point because that is something that wasn't clear from the Meta announcement either.

It was saying, okay, we're gonna train people as electricians and welders and HVAC. I forgot what the last one was, but you know, very much skilled trade jobs, but it wasn't necessarily, oh, we're gonna train... You know, I sort of assumed, but probably erroneously, that it's like, oh, if you're training as an electrician because of a Meta program in a place where there's a data center, it's probably like data center electrician, like be electrician for our AI data center and understanding how to use AI in your job.

But not necessarily, right? It just might be doing the kind of electrical work that's needed in a data center and not using any AI in your job and being just an electrician that's specialized in a particular area. Still might be a steady job in an era where there's lots of data centers, but it doesn't necessarily actually give you the AI skills to be able to translate that into something else, which is a really big distinction.

I wonder how much AI training is in this kind of blue collar training. It's, it's gonna be really interesting to see it play out. 

[00:15:28] Gemma Lenowitz Picot: Yeah. And I'm optimistic that entrepreneurship is a pathway for a lot of these skilled workers, and there's the opportunity to think about the role of the union to create some bargaining power at the system level, but also the role of AI to help small business owners create a guild of welders, for example, and to create opportunities for on-ramps and apprenticeships into that industry for folks that might not otherwise have a union path, which is not how these data centers are being built out at the moment.

[00:15:59] Alex Sarlin: So we also saw an interesting, you know, speaking of the IPOs and things that are coming, we saw the Claude Core announcement was kind of a, of a different flavor, which is not necessarily blue collar training as much as trying to extend Claude usage to nonprofits. So they said, you know, at least 400 nonprofit organizations are gonna get, you know, $10,000 grant and free credits to use Claude and training, and they're basically trying to say, "Oh, you know, yes, we know these giant companies, the Walmarts and Amazons of the world, are using huge amounts of tokens, and they're doing all sorts of AI.

But if you're a small nonprofit that's doing really important, meaningful work, you should also be able to use Claude Code and Claude Cowork and to do really amazing things." And they're trying to sort of extend capability. Again, do you see this as pure, you know, they call it beneficial deployments at Anthropic.

Is this, is this pure good, or is it a little bit of a, "Hey, we're just trying to extend to another sector," or, or both? 

[00:16:55] Gemma Lenowitz Picot: I take seriously Anthropic's claims that they're trying to ensure AI's benefits reach civil society more broadly, and I think this is an important step in that direction. It also trusts local expertise, community expertise about what is needed because these nonprofits are so often rooted more deeply in community than, you know, a frontier lab might be able to understand the nuance So, uh, what I like about this program is that they're combining some training for the fellows with paying the fellows' salaries and of course giving them some credits to use Claude.

I hope the fellows will also, you know, apply other AI tools where possible. But I think this is both a workforce program and a play to drive AI adoption infrastructure in these nonprofits. And there are a number of organizations working to expand nonprofit capacity. What's interesting is that some of them are routing folks to courses like the ones we talked about earlier on OpenAI Academy, focused on the vertical, do you work in sales or marketing or fundraising?

But I think this is ultimately about building more capacity in these organizations that would otherwise be unable to fund and resource the time behind training workers. Whether this creates a pipeline for, like, first gen workers, I'm not sure. I think they are open to non-degree holders and, you know, recent college grads.

And I hope that the core also involves some kind of cohort-based learning. So I think worth watching whether this becomes a model that others replicate because I think the reception so far has been quite positive. 

[00:18:24] Alex Sarlin: Yeah. It does feel different sort of in, not just degree, but, like, in type from some of the other announcements we're seeing where that are sort of tied to...

that are blue collar, or they're tied to data centers. This feels a little more like it's a sort of national program that's really designed to sort of regionalize and support the types of organizations that often aren't at cutting-edge tech even before the AI era. Non- often nonprofits didn't always use the most cutting-edge technology, or they weren't able to hire super experienced programmers, you know, in the past.

And this is a really interesting opportunity for nonprofits to actually have a little bit of like a leapfrog moment where if, you know, if you're a small nonprofit with 10 people, but you can use Claude code incredibly effectively, then you could maybe have the capabilities similar to a much larger organization that can actually hire a full engineering team, which is-- this has been, you know, traditionally a real bottleneck for nonprofits.

They just often can't afford certainly AI engineers, but, but let alone, you know, traditional web app engineers. So it is a really interesting play or interesting idea, and hopefully it could have all these downstream effects, both on the type of training, to your point, but also on the capabilities of nonprofits.

Like, it would be amazing to see all these local nonprofits suddenly have, like, spectacular websites and mobile apps and being able to do really great marketing and fundraising and being able to, like, create... I mean, that would be, that would be a beautiful, beautiful outcome of this. 

[00:19:42] Gemma Lenowitz Picot: Agreed. And this is Anthropic really putting trust in CodePath as their signature partner around economic mobility.

And CodePath has experience placing people in tech jobs, but those are, are tech roles which are sometimes at banks or big tech companies and, you know, we really should not Understate the significance of that first rung of the career ladder being removed. And so this is also about finding a home for those talented and motivated young people, and hopefully this is an opportunity to contribute some good.

I hope also, and this is an opportunity to get on my soapbox and talk about how many nonprofits we have in the US. It's too many. I think it's around a million, and so registered 990s. Don't quote me on that, but I think the answer is too many. The opportunity now to also use technology to help organizations think about mergers and acquisitions.

I think there's a lot of dialogue in private equity and social enterprise spaces about organizations that are swimming in the same direction and could strengthen by joining forces. And so hopefully this also is an opportunity for Anthropic to map out which kinds of organizations are doing what kind of beneficial deployments, and then maybe supporting some of the connective tissue between those, those groups 

[00:20:52] Alex Sarlin: That's a very interesting point.

That is a large number, or even if it's anywhere close to that, that's a lot more than I would've expected for nonprofits. You know that space more deeply than I do, and that is a really interesting point that you have all these very small nonprofits. Some of them are working on similar problems or working in similar ways.

Maybe they'd be more effective if they merged and went together, and then if you combine that with technical capacity, you could really increase their impact. That's exciting to see. Y- you know, this was a smaller story, but it was interesting because, you know, we, we're talking about some of the big frontier labs.

But, you know, we should say in passing another thing that came out this week, but it's-- I, I, I don't even know how relevant to EdTech it is, is that we're-- there's just been this chaos with Anthropic rolling out their most powerful model, Fable, and then the White House coming down and saying, "No, we wanna make sure that foreign nationals don't have access to this," and having this sort of national security internal panic, which then led Anthropic to have to sort of pull it back from everybody, and there's all sorts of kerfuffle happening there.

I don't know if you wanna talk about that for a little bit. I don't know how relevant it is to EdTech. Other than that, once they're on the other side of this, Fable is an unbelievably powerful model. It can do really amazing things. I've seen a lot of people building educational games, for example, with Fable, that it's much easier than any other model.

I'm curious what you make of that. Where is that gonna land, and is this just a passing story, or is this just gonna be a government regulation and AI starting to really butt heads? 

[00:22:15] Gemma Lenowitz Picot: Yeah. It's hard not to tangle this with the valuation narrative going into the IPO as well. There's- 

[00:22:21] Alex Sarlin: Mm. -

[00:22:21] Gemma Lenowitz Picot: some skepticism around the claims that Anthropic is making, right?

It's, it's certainly in their best interest to market this as a transformative and dangerous technology if that doom leads to some investor frenzy around the opening share price. But I, I guess the connection to EdTech is around risk assessment, and I think we in EdTech hope that school districts, teachers have the opportunity to select what works best for their students.

And risk assessment has historically lived in IT departments. I think AI is really introducing a lot of uncertainty around how fast to move, knowing that there's urgency behind our, you know, struggling outcomes, and the idea that we should patch vulnerabilities. I think, you know, the other aspect of this model is the sort of cybersecurity implications, and certainly we've seen that with PowerSchool and other players in education.

[00:23:16] Alex Sarlin: And Structure. 

[00:23:17] Gemma Lenowitz Picot: And so, yeah, yeah. So we need to kind of figure out-- and in this case, maybe there is some, like, strategic alarmism about insisting that this is dangerous, but I think hopefully our field moves in the direction of advancing data privacy and understanding with more fulsome effort where the vulnerabilities lie because, you know, I certainly don't want to see another disruption like PowerSchool and Structure.

[00:23:41] Alex Sarlin: Hearing you talk about it brings up an idea for me that we had talked about a lot for a while on this show, but it, it sort of has gone under the radar, but maybe should come back up, you know. There was a while when, when, when Google was first putting out its LearnLM model. It was basically training a model to be really, you know, sort of learning-focused, to be designed for learning outcomes, to try to be, you know, sort of a safer model.

They were thinking about LearnLM as sort of a separately trained model, and then what they ended up doing was sort of feeding LearnLM back into the core Gemini models and some of the thinking into it to try to sort of inject that learning philosophy into their core models, which, you know, there's a lot to be said for that.

But at the time, we talked a lot about the idea of like, well, what if these frontier labs had sort of education-specific models? Because when you think about something like Fable, Mythos/Fable, it's like as the capabilities of these models just get sort of like almost like mind-blowingly bigger and they put out these things like, "Hey, we don't trust the world with Mythos because everybody can use it to hack everybody."

You know, it's like as these things get more powerful, they get more unpredictable, they give people more capability to sort of cause chaos, and it's not really a safe environment if you have all of these EdTech companies already in schools, including big ones like, you know, incumbents, but also startups, and they're using these underlying models, and then the models are getting more powerful and can do more and more.

If you're a administrator, you're like, "Am I gonna give a tool to my teachers and students that allows them to like hack the bank?" You know, like, I mean, it's just, it's a really weird thing to think about. But if there was an education-specific sort of branch of these models that was obviously designed for learning but also designed for privacy, designed for security, designed not to build its power so quickly as, as the commercial models, the commercial core models are, I think that could be a real win-win.

And, um, like even at the higher education level, right? If Anthropic had something called, you know, Claude Scholar, or I mean, that's Google Scholar, but like Claude College, that basically it was the university framed model and it was incredibly powerful for research, incredibly good at teaching, but was not designed for hacking.

It was not designed for collecting data, and you couldn't do some of the, the dangerous things. Like, wouldn't that be better for Anthropic but also better for the colleges that use it? Like, I wonder if that's something that is on any of these frontier labs' minds. I'm curious how you react to that idea in general.

[00:25:55] Gemma Lenowitz Picot: I'm sure they would love to solve it. And I always come back to the idea that like Google solved this for search in the era where, you know, it was certainly possible that folks were typing in how to do destructive things and get, you know, recipes or playbooks for those. So we don't have the same capability of controlling LLM output.

It's not a deterministic system. But I think there are a lot of layers and wrappers that you could use to improve- the sort of like beneficial outcome of the output. And so I think that this kind of overlay is why OpenAI put an early bet on Khan Academy and to build out Khanmigo. We can talk about now how folks are getting their recommendations for AI tools from AI with the new EdTech AI Visibility Index rating.

And, and, and Khan Academy was on that list with, I think Duolingo and Coursera. But those are strongly well-known brands, I'd say. And part of that brand equity comes from being a secure and reliable tool And, you know, I think philanthropy certainly suffers the shiny new object problem. But I think in, in learning and teaching, we often also suffer a shiny new object problem and, and the desire to innovate without slowing down to understand if something really suits the context properly.

But I think putting more trust in organizations that are taking the time to build safe products is important, and, and hopefully parents are-- and families are making the same decisions about overlaying some kind of quality assessment in their process of decision-making. 

[00:27:23] Alex Sarlin: Hundred percent. Quality assessment and, and safety protocols.

And I mean, the question is just sort of where on the stack that happens, right? Does it happen all the way at the base model level? Is, is the model itself trying to do, as you say, what Google Search does and say, "Hey, if somebody asks how to make a Molotov cocktail, I'm not gonna answer that." Or is that actually in the edtech tool that is being built as a structured delivery system on top of a wrap?

Or does it happen all the way at the school level where they have to be able to set constraints? Obviously, for Molotov cocktails, that wouldn't be-- it wouldn't be that deep. But like for, for certain types of searches, cer-certain types of questions. I think it's like a core question of the age right now for edtech is like sort of at what level of, of the tech stack is the education and the safety happening.

Is it happening all the way down or, or at the top? S-so you mentioned this, this re-- This was a really interesting article, and we should definitely-- This will be in the, in the links for this episode. But there's a company called 5W AI Intelligence that sort of does industry information management and puts out papers and things, and they, they did something called an Edtech AI Visibility Index.

They basically said who in the edtech AI space is most visible, who's being cited the most, who's being sort of noted and searched for. And this is a really interesting list. I don't-- It's-- In some ways, it's not a super surprise, but there are some surprises on here. I would love to actually go through, uh, like at least the top 10 on this list 'cause I think it's-- it'll be inter- This is an edtech audience.

This is like-- We're-- This is exactly the right group. So you mentioned number one on this list is Khanmigo, right? When you're talking about edtech AI, the most visible product in edtech AI, according to this survey, is Khanmigo, even though the Khanmigo folks have come out and been like, "We're still figuring this out a little bit," right?

I mean, they have not declared victory and said Khanmigo is the AI tutor that solved the problem. They're like, "We are-- We're still working on it. And, and, and it can do amazing things, but there's still a lot to learn." Duolingo is second, not a surprise. Mostly consumer. Duolingo shut down its schools product.

Coursera, Magic School, Udemy, Masterclass, Chegg, Synthesis Tutor, interesting one. Khan TED Institute is number nine. That is- 

[00:29:23] Gemma Lenowitz Picot: Brand new. Yeah. 

[00:29:25] Alex Sarlin: Brand new, and it's already i-in the top 10. That's really interesting but obviously related to Khan. And then Course Hero, you know, slash Lernio at number 10. In some ways, it makes sense.

These are, these are some of the most visible brand names in education, you know, bar none. I'm a little surprised to see Masterclass on there because they haven't actually focused on AI as sort of a core thing. MasterClass has just launched this incredible executive ed program that is AI-focused. I actually am, am gonna be part of that in, in a few weeks.

I'm excited about it. I'm gonna be there as a student. But I wouldn't think of MasterClass as pure AI, especially compared to places like Khanmigo or Magic School or Synthesis even. What do you make of that list? Does this seem like, hey, that makes sense, or anybody surprise you on there? 

[00:30:05] Gemma Lenowitz Picot: Well, first please report back on your course experience on the pod 'cause we'd all love to hear about that.

I 

[00:30:09] Alex Sarlin: will. 

[00:30:10] Gemma Lenowitz Picot: But I think this comes back to this idea of GEO and what the winners are going to be AI legible or LLM endorsed if they have a good content strategy, and that's like how much do you write and publish and structure cleanly the information about your product, and that will probably determine reach as much as product quality.

Certainly this is a whole category being developed as-- of marketing firms that are helping folks perform well on LLM search. But it remains to be seen if there's some kind of quality overlay you could add, and ideally the LLM would prompt you back with more, you know, to request more specificity on what you're trying to do.

Yes. But I think most people are typing a really quick like best summer camp for fifth grade or, you know, best math workbook for calculus or something, and not necessarily getting into the, the detail that would be required to make a specific recommendation. 

[00:31:02] Alex Sarlin: That's a great point. I mean, we've mentioned on this show for a long time that I think that the only two EdTech companies that have truly become like household names are Khan Academy and Duolingo, maybe Pearson if you consider it an EdTech, and maybe Coursera at this point, I think it, it-- that's number three on this list.

It makes sense to me that those sort of take the top three spots of Khanmigo, Duolingo, and Coursera. Yeah, I mean, nothing on here like super surprises me. I'm sur- I am a little surprised how high Synthesis Tutor is. We don't talk about that, that company that much. I'm a little surprised how high Chegg is, but, but it also makes sense.

Chegg has been wrapped up in this AI debate for a long time, and it's been in enormous trouble because of it, but it is, it is, it has gone-- it went AI very early, and it kind of makes sense. But it's interesting and I, I lo- I mean, I hope there's actually more articles like this 'cause I don't see much about EdTech visibility.

You know, we talk about it here, but it's interesting. There's also a few consumer plays on this, right? I mean, you look at Duolingo, Udemy, MasterClass, Synthesis Tutor And even Khan TED, I think is going to be, is, is basically cons-- Well, you could look at that either way. But a lot of these are consumer products rather than products that sell to schools.

What do you make of that? 

[00:32:09] Gemma Lenowitz Picot: Yeah, I think that just shows who's doing more of the search, right? You don't have district administrators typing in what to procure because they know they're gonna have to go through a lengthy process anyway or issue an RFP. But I think consumers are the ones out there searching, you know, how to learn Spanish or how to get a degree.

And some of it is probably workforce motivated, right? We know in this labor market, people are being creative about finding pathways into job opportunities. And so I bet some of it is driven by, especially the, you know, new TED-Khan Institute is probably driven by folks who are looking for low-cost degrees and, and pathways to higher paying jobs.

Just the other piece I'd flag is that districts are figuring out if they're going to pay for the things that used to be free And so there may be more consumer swing if schools decide to drop a whole slew of products that had some kind of consumer appeal, and maybe students and parents are willing to pay, you know, a small amount.

But the question is also like how do you price that as a, a tech entrepreneur knowing that your costs might be kind of lumpy moving forward and for tokens to get more expensive? 

[00:33:12] Alex Sarlin: Yeah. That would also be an interesting use case for an education-specific AI, right? Because if, if you had a model that was designed for a learning use case, then you could actually relatively contain the token cost and not have it change enormously based on what model you're using or how complex the ask is, or whether you're on Fable or Opus or Haiku or all those things.

But, but I hear you. And who's doing the searching is, is key here. I, I, it's always been interesting how different the consumer edtech and the B2B edtech sectors are. When people come asking about-- come to me and ask about what, you know, if they're trying to enter the field or they're trying to understand the field, uh, one of the first things I always say is like, "Well, B2C and B2B are like hugely different in edtech.

They just are very, very different worlds." There are people who've crossed over, there are people who do both, but like, boy, are they different. It's interesting when you think about in the AI world how some of the, you know, things like Synthesis Tutor, which is totally consumer focused as far as I know, I don't think they have a school product, can grow a lot sort of under the radar.

Or there's a lot of AI apps. Uh, we interviewed a, a founder last week who do-- who runs a company called Curiosity, has an app called Ace Tutor. It's like one of the top apps in the, in the App Store for AI learning, but we don't think about it that much because it doesn't sell to schools, it doesn't have that much vis-visibility except directly to consumers.

It's re- and there are a whole bunch of those. It's really interesting how that works. One story that caught both of our eyes, I think you really s-spotted this one, which is really cool, is that there's an article in The 74 this week about how on DonorsChoose, the requests for AI-related tools have tripled in the last couple of years, which I guess makes sense, but I wouldn't have necessarily guessed that because I-- you often hear about the backlash against AI and teachers not wanting to use it or being concerned about it.

Instead, a lot of them are asking about it, especially for students with disabilities, for English language learners, for underserved students, for things that do translation or, or IE- IEP support. That was a really interesting and sort of a uplifting take in a sort of dark time for edtech AI. It's like teachers are actually saying, "Oh, I have all these students who speak different languages.

I wanted translation pens for them that use AI to support them." Like, that's wonderful. What did you make of that story? 

[00:35:18] Gemma Lenowitz Picot: Yeah. I'm glad you're feeling optimistic about it. I always get a little bit, little bit saddened when I hear that teachers are requesting things that ideally schools should be able to provide- Of course

at minimum. But I think this is about teacher voice, and so I'll take the optimistic case and say, you know, that these teachers have decided there's a unique product that works best for their students. So they still have to go through the, um, district procurement process, by the way. If they procure any software through DonorsChoose, it still has to follow All of their protocols.

But these tools are, you know, I think the idea that, like, teachers also want agency, and the idea of waiting for top-down decisions doesn't always meet their student needs. And so this is about using AI as an equity lever, which is certainly not something we hear a lot about the frontier labs. And th- these teachers are trying to close gaps for the students that are most likely to fall behind.

And I think in some of these products, you really see engagement, right? Like a translation pen or a personalized tutoring software is really about bringing students back into the classroom and, and getting them into the zone of proximal development. And I think in a tighter fiscal environment with schools, you will see less of the bottom-up motion with EdTech sales But this signals that teacher-driven deployment may be a real adoption path.

You know, I certainly wouldn't like place a marketing bet on DonorsChoose as a company, but I'm heartened to see that, you know, in light of all of the learning loss that happened from the pandemic and, and also shrinking demographics being a, a real headwind for schools, that this is an opportunity for AI to slot into strategy and to improve sort of like teachers' ability to meet their students where they are.

[00:36:52] Alex Sarlin: Yes. And you know, I, I connected to the announcement a few weeks ago from Randi Weingarten and the AFT where they started-- They basically, you know, swapped their position and said, "We need to really keep AI out of schools, especially at the elementary school level." And one of the comments in that, in the Times article about that was, well, some teachers are saying, "Well, this undermines teacher autonomy," right?

If the union is saying, and is saying, you know, "Our official position is no AI for elementary school students," but you have teachers who really want these translation pens for their students who speak all these different languages or really want an engaging AI tutor for the students who are falling behind and they don't have time to remediate or don't have the ability to remediate.

Like, I feel like that's exactly where the rubber meets the road, right? Where y- you know, you can have the, the whole field getting hot and cold on AI, but when you're a frontline teacher, when you're absolutely-- when you're in a classroom and you have the student who just came here from X country, from Bulgaria and speaks Bulgarian and l- and doesn't know what's going on, and you're like, "There-- I know for a fact there are these tools that can make this kid's life so much better," but like, you know, the tectonic shifts on AI and how people feel about it are, are sort of closing off my options.

I don't know. It feels like that bottoms-up motion may be a little bit of a mitigation for some of the school district union policies that are, that are increasingly, you know, trying to shut down some of the AI and screen time general policies. So it's, it's, it's gonna be interesting to see how this plays out.

But this, this article I should mention, was by the CEO of Do-DonorsChoose. So it's-- this is the DonorsChoose CEO talking about how on DonorsChoose there's some really interesting stuff happening in AI, and it is equity-based. But yeah. 

[00:38:26] Gemma Lenowitz Picot: Do you see this as connected to the micro-schools movement or ESAs or school choice?

This is, you know, this is also m- ultimately decentralizing decision-making. 

[00:38:35] Alex Sarlin: Exactly. I mean, I do. I have been worried for quite a long time, from the very beginning of when, when some of the big school systems, New York and, uh, Baltimore, I believe, started banning AI, right? You know, sort of knee-jerk bans right off the bat I've been really worried that what's gonna happen is that if public...

I mean, this is sort of a, a microcosm of the school system at large, right? If public schools, because of these large, complex political issues and because there's so many people involved and so much training involved, if the public schools start to turn away from this technology and say, "You know what? We just don't see a place for AI.

It's too risky. It's too complex. We don't wanna deal with it. We don't wanna do the training. We don't wanna worry about tokens. We don't wanna worry about breaches," and they say no, then what you're gonna get is more and more parents who, the set of parents who say, "That's not a good decision. AI is clearly part of the world now," pulling their kids out of public school, going to micro schools, going to charters, going to places like Alpha School that are, like, very, very aggressively AI focused or, or Acton Academy or y- a bunch of different flourish schools.

There's a whole bunch of things happening there. And I think that, you know, if I was the head of Alpha S- if I was, you know, McKenzie Price right now and I see this happening in regular public schools, I'm happy because I'm thinking, "Okay, public schools are now going back to this crouch and defend position and not feeling like they wanna be out front and not feeling like they wanna be innovative and starting to say, 'This is just too scary.'"

And we see the political winds and people are getting upset about because of the Jonathan Haidt book or because of the, the Horvath book, people are thinking that edtech and AI are no good anymore, we're gonna shut this down. They're gonna say, "Great for us," because so many parents are gonna say, "They just shut down AI in my public school.

Now I'm, I'm gonna s- consider spending a lot more money for a AI education because it's just not gonna happen in, in my traditional school." And I think you'll have tons of, of people replicating that model. So I think it's super relevant to micro schools. I don't know, what do, what do you think? 

[00:40:33] Gemma Lenowitz Picot: I agree. I just wish there were more opportunities to give parents, you know, voice and input into what's happening at their public schools.

And I think the whole future of our democracy sometimes feels like it's in question. And so I'd like to see, you know, lots of opportunities for teachers and families to align on what students need and, and to provide some input, and also for quality to be, you know, unnecessary. And ideally we have a stronger research base for any intervention that gets rolled out.

[00:41:00] Alex Sarlin: Yes. 

[00:41:01] Gemma Lenowitz Picot: But of course, there's so much newness with AI. This idea of backlash and screen time and the sort of science of screen time reminds me of what's happening across the pond. So apart from a 4:00 p.m. kickoff against Croatia for the English men's World Cup team, there have been a, a few big pieces of news out of the UK.

[00:41:20] Alex Sarlin: Yeah. Let's, uh, go for it. 

[00:41:22] Gemma Lenowitz Picot: This ban on social media for children under 16, which I think partly was inspired by Australia's move late last year to do the same thing. So this ban covers Snapchat, TikTok, Instagram, YouTube, Facebook, and X But not for the messaging services within those platforms and not for YouTube Kids, which is, I guess you could argue a separate product.

But they are trying to bring regulation to the British Parliament this year and target to begin this in spring of '27. And so, of course, the backlash to this news, to the backlash news, is all of the advertising revenue. And I was really shocked to learn how much of the platform revenue from these social media companies comes from the under-18 market.

For Snapchat, it's 41% of their ad revenue. That was upsetting to learn. And Instagram has, you know, accrued billions of dollars in revenue from ads targeting minors. And so, you know, this is a, a little bit of a question for some of these US-owned and managed tech companies to respond to, but I'm curious for your take on, you know, does social media-- Well, I guess, does a ban work?

Are students always gonna find a clever way around it? And then how might this affect that broader conversation about screen time? 

[00:42:37] Alex Sarlin: Yeah. Great question. So this is a really nuanced and complex issue. I just let me say that right off the bat, but I, I say that one thing I really like about this, so w- we should say that this is paired with, so the UK is, is, is considering this super intense social media ban.

They're also trying to expand AI tutor usage for low-income students. Basically, students who get free meals at schools, you know, which is a famous proxy for underprivileged students. They're basically promising AI tutors for almost half a million students in the UK on free school lunches. Now, think about the combination of those two stories together, right?

So anxious generation comes out. All this research starts to be very clear that social media has some pretty serious negative effects on students' mental health, right? We- we've seen depression, we've seen suicidal ideation, we've seen anxiety spike like crazy, and people are starting to say, "Hey, this all happened right when phones and right when social media started to go everywhere."

In the US, we're seeing this weird conflation of those things, I think. It's a weird conflation where they're saying tech is bad. They're saying, "Oh, social media is bad. Well, so is edtech. Edtech's bad too, and we should get tech out of schools, and it's the screens, it's the phones, it's the devices, it's the laptops."

I don't think that's actually the right response to this. I think the right response to this is a lot closer to what you're seeing Keir Starmer say, which is this combination of, we've got to get social media out of kids' hands because it's really hurting them in very quantifiable ways. Yet, educational technology is very powerful and can, if used correctly and designed correctly, have very good outcomes.

Why don't we actually try to figure that out? They're piloting these AI tutor programs. Obviously, gonna check the, the evidence on it before scaling it, but that's the plan as I understand it. But like, I think that's a much more nuanced and smart reaction to the tech backlash than just saying, "Technology is bad.

We need to go back to blue books and handwriting and paper and pen." I like-- I think that's very strange. That doesn't actually answer your question about whether this will work. I mean, I've seen some interesting things online about, you know, students being in very aggressive rebellion against soc- these social media bans.

They are furious, as you can imagine. But I think this kind of like tough love, like, "Hey, look, let's, let's protect..." It's, it's paternalistic, right? Like, let's protect our young people from these profit-driven companies like the Snaps of the world that, that may be really harming them, and we have more and more evidence that it's harming them like nicotine, right?

Like, like that kind of paternalistic approach I actually am in favor of. When it expands, I think sort of blindly to all technology, that just seems ridiculous to me, and that's what I'm afraid has been happening in the US. I, I don't know. What is that reaction? How do you see it? 

[00:45:25] Gemma Lenowitz Picot: Yeah, I agree, and, and that I-- this sort of blunt instrument of bans can be really tricky, and, and age limits also matter a lot here I think that the idea of the sort of inherent value of social media is pretty low for kids.

[00:45:40] Alex Sarlin: Yeah. 

[00:45:40] Gemma Lenowitz Picot: And edutainment now rises as a category if you can't have social media. I've been wondering going into the summer, is there going to be a new, like a cool new thing that young people are using a la Roblox or Minecraft or I don't know that it's Khanmigo, but wouldn't it be great? I think the evidence for AI tutoring is nascent, so it's interesting to hear the sort of like optimism around it.

You know, I think a lot of the US researchers would say you really need humans. It's the human relationship that drives the quality of the tutoring, and Susanna Loeb taught me 5% of students have the self-regulation and executive function skills to engage with the standalone AI tutor as intended. And so, you know, hopefully, the UK will also use this like scaled commitment to AI tutoring to build a better implementation playbook.

And if the kids can't access social media on the side, like maybe that will help that implementation come to life. But yeah, I don't know if the lobbies of the social media companies are so strong in the US, you know, if it is ridiculous to expect that we could ever see a parallel process here. But I think, you know, parents are certainly coming to some awareness, and hopefully EdTech can continue to drive understanding of the value of, of products that might include screen time, but ultimately also drive learning outcomes or social emotional outcomes.

[00:46:59] Alex Sarlin: Very well put. And I, I, I, I agree with you that I think just providing access to standalone children accessible, like instigated by kids, AI tutors that don't have a human... They don't have structure around them, they don't have humans in, you know, interesting new papers about relational intensity or va-various things that need to have be designed into the system that have actually, you know, humans with AI working together.

Totally on the same page with you that I'm not trying to say that, you know, standalone AI tutors are the answer. I don't think they are by themselves. So yes, you need a system. But I, I just... I like the idea of saying social media, I mean, the 16 is no coincidence, right? That's exactly the recommendation that, that Jonathan Haidt gives in, in, in the book, and that what a lot of is becoming sort of the consensus of...

You know, so, so like, you know, the idea being like we need to get kids away from the social media that's hurting them and try to provide meaningful education technology that's helping them. I just think respect that those are being said into, you know, that those are going in different directions.

Whereas here, I think, you know, yes, the lobbies are way too str-- I mean, it's not even just the lobbies. And we, we have a, we have a moment in American politics where it's like there's very little appetite for social media regulation in the US at the, certainly at the federal level. I think there is appetite at the state level, and I think some states are probably heading there actually.

But it's very hard to ban social media here. But conceptually, I respect the idea of saying, at this moment where we're sort of having this reckoning of the last major move in technology, which is social media, and seeing that it may have caused much more harm than good, especially for young people, being like, "Okay, we have to deal with that," but let's not let that, you know, throw out the baby with the bathwater and think that all technology is negative, which there is not a...

The evidence is nascent for a lot of this, but very little of it says it's harmful. Some people may disagree with that. They're looking at cognitive Outsourcing and offloading and all that, there, there is some concern. There are some studies coming out there saying that, you know, people are, are using AI to avoid learning, so I can definitely see that.

But, but most of the EdTech AI progr- AI products, it's-- even though it's early, I think there's still potential there. We are not, you know, 10 years into it and saying that the kids who are exposed to AI tutors are miserable and none of them succeeded in school like the same way we're seeing that real, real serious evidence of social media.

But I hear you. I don't wanna weigh too hard in on the AI tutoring piece because it is-- we're in a really interesting moment for that. 

[00:49:22] Gemma Lenowitz Picot: Agree. 

[00:49:23] Alex Sarlin: But yeah, I still think there's potential there. I don't think it'll be pure AI tutors that are pushed, you know, or whatever, pull only that are the, the 5% problem is very, very real, and we've seen that over and over again w-with, with AI tutors and even pre-AI tutors with, with just online tutoring or access to human tutors that is instigated by the student.

Doesn't work. That did not work during the pandemic. So I'm with you. You need a system, you need a structure, you need it to be instigated from somebody other than the student. I think that's kind of the core thing. Anyway, lots and lots unpack there. I feel like I'm on my black heel there. One more story. I know we're sort of running along here, but one more story that jumped out, and I'd love to hear you talk about it, is the Microsoft CEO, Satya Nadella, had a really interesting editorial, like a LinkedIn editorial, about this concept of AI learning loops and how that's gonna frame the future of workforce.

Tell us about it and, and why this stood out as something that might be interesting for the EdTech community. 

[00:50:18] Gemma Lenowitz Picot: Yeah. I think what Nadella did in this piece is make the case for a new measurement of token capital. So he describes learning loops as a system where employee judgment and sort of quality assurance process and upskilling with expertise kind of continually improves internal AI processes, and that creates token capital.

So a form of AI capability that an organization might hold on a balance sheet like they do other kinds of capital. I think this is also trying to put enterprise AI usage in the context of human capital and understanding how value flows to the models versus to the humans in an organization. But this is really a, an opportunity to think about organizational learning for enterprises.

I'm curious how this plays out for, like, public institutions like nonprofits and schools, because part of his argument is that if you outsource all of your AI learning to your vendor, then you don't build internal feedback loops. And this is so much about the change management that has to happen in organizations to benefit the most from AI.

So not using it as a crutch or a task completion tool, but really using it to enhance learning and system improvement. And this is to your point about cognitive offloading. Like, I think that's not just happening with young people, it's happening with adults. And so how do we ensure that enterprise employees are also building learning loops that become kind of the IP and an asset of an organization?

This idea that any company can basically buy a subscription to a Frontier Lab model, but you can also build a harness or you can build on top of it and have all kinds of proprietary information and accumulated expertise internally. And I think we're moving from an era where the organizations had a few, a smaller number of AI champions to an era where there is more mass adoption in enterprise context.

And so I'm hoping that this means people will be thoughtful about this idea of token capital as it relates to human capital, and for all organizations to kind of upskill in AI toward achieving their mission. 

[00:52:23] Alex Sarlin: Yeah. It's a really interesting idea, and that, that loop sort of is back and forth between what he calls the human capital employee knowledge and employee ability, uh, you know, the workflow, the understanding of what's important in a, in a business, and the AI capability, which is the sort of proprietary set of data and workflows and evaluations within a company.

So it's not, it's not within, at the model level, right? It's not like Claude builds a, a workflow for Boeing, and then the Boeing employees are giving feedback to Claude. It's that the Boeing employees are giving feedback and creating a loop with the AI systems within Boeing, which then improve the understanding of the humans, which then improve the understanding of the AI within Boeing, and that's what he sort of means by the learning loop.

It's a really interesting framing, and I love your point about how this might work in some-something like a school or a university. I mean, 'cause absolutely the same dynamic could apply there. And, and as people are getting more thoughtful about token spending, the idea of token capital meaning like what's your ROI on every token you spend, is it actually, you know, or every million tokens, whatever it is, it's like, are you actually getting value out of it?

Well, that needs to be evaluated by humans, and I think that is part of that learning loop as well. It's a really interesting framing. Uh, we will put a link to that article and the original LinkedIn post in the show notes for this episode as always. This has been so much fun, Gemma. You bring so much expertise in so many different areas than I have, and, and certainly than I'm, I'm used to on this show, so I'm really, really had a ton of fun talking to you today.

We're gonna move to our guest segments right now . Thanks so much for being here with us. Thank you, Gemma Picot from Samvid Ventures. If it happens in EdTech, you'll hear about it here on EdTech Insiders. 

[00:54:00] Gemma Lenowitz Picot: Thanks, Alex. Good to be with you. 

[00:54:02] Alex Sarlin: We have a fantastic guest on the Week in EdTech at EdTech Insiders this week.

We are speaking to Tom ap Simon, who is the president of Pearson Higher Education and Virtual Learning, leading teams that serve millions of higher education and K-12 learners globally, including Pearson's career readiness programs. We're definitely gonna talk about that. With a focus on digital product innovation, Tom is at the forefront of Pearson's work to enhance learning with AI tools.

Tom ap Simon, welcome to EdTech Insiders. 

[00:54:33] Tom ap Simon: Oh, thank you very much, Alex. It's great to be here, and I'm really looking forward to the conversation. 

[00:54:36] Alex Sarlin: Me too. So first off, Pearson is a giant in edtech, and it is a giant in career readiness training in higher education. Tell us an overview of what is top of mind for you as the president of higher ed and virtual learning in 2026.

[00:54:51] Tom ap Simon: Yeah, it's a great question, Alex, and I think it's a really fascinating time to be in the industry because there's so much change and yet so much constancy, right? You know, there's so much hype about AI, yet the reality is we're seeing that education is still struggling with many of the same old issues.

How do we get kids to pass Algebra 1 in ninth grade? How do we get people set up to be successful and actually have a clear understanding of what career choices they've got? Thinking through whether two year or four year is right for them, or whether a, a skilled trade is the right option. So there are-- there's all this hype, and then there's all, like, the real world.

So it's trying to distill some of the hype from the real world, and it's working out fundamentally, I think, how we can really help kids make better choices that will enable them to be more successful. And also underpinning all of that, it's really around outcomes. Of course, we talk about AI because everyone's talking about AI right now.

But really it is around how do we get kids into the right roles for them, into the right jobs for them to make them successful regardless of, of where their background is, what their interests are. And so we're thinking a lot about that in terms of positioning children and, you know, young adults and working adults to be successful.

So in many ways, the fundamentals haven't changed, but it's about thinking about those very differently now in this new sort of paradigm that we're working in. 

[00:56:04] Alex Sarlin: I love that point about, you know, the challenges that have existed in education for a long time, the algebra one, the career exploration. You know, a lot of things that have been missing for a long time are still missing, and yet we have this whole new set of challenges and opportunities with AI that are just sort of added to the mix, making it an even more complex picture.

We're at this funny moment, you mentioned the sort of hype and the reality not yet matching up. AI tools are more accessible than ever. They're more used than ever. In the last three years, we've seen, you know, majorities of students, majorities of educators start using them, and students at the higher education level are using them absolutely everywhere.

At the same time, employers are still saying that graduates are not necessarily ready for work in the AI era or, or, or before it, but let's say in the AI era. So what do you think is the disconnect? People are using the tools more than ever before, but they're not yet being considered skilled enough to jump right into the workplace.

What could we do here from an edtech and an education perspective? 

[00:57:03] Tom ap Simon: Yeah, it's a great question, and it's a funny one, right? Because of course, if you believed all the hype, then, you know, we would have no educational issues whatsoever. But of course, the world is very different. I think fundamentally kids or people are not using AI tools in the right way from a learning perspective, and so therefore, they're not necessarily learning how to learn or how to study and how to learn.

And so that makes it hard for them to succeed at work. And so if you're not learning the practical skills needed in the workforce, then you don't really have the job-ready skills that our employers are looking for. And so I think a lot of the student use today is still sort of informal, and it's disconnected from how learning is actually designed.

So like you have widespread tool usage, but sort of incorporating AI thoughtfully and deliberately into learning is still very, very uneven. So that applies to professional development, that applies to, you know, how s- instructors are setting up their courses. And so you've got students using a tool from time to time or all the time, but not necessarily having the right judgment, the right context, the right environment to use that tool properly.

And that then means that learning becomes uneven, and it creates a real difference in terms of outcomes. And one of the things that our recent study showed us was that 92% of students are

using AI tools, yet only 14% feel their education has set them up to be successful. That's a shockingly large disconnect, and so therefore exposure alone isn't enough. And so I think that says to me that we have to sort of respond both by r- supporting instructors, instructional design, faculty, uh, they need more support, they need more ability to be set up and to be successful, and that's easier said than done.

That requires deliberate investment in faculty professional development, and we know that budgets are tight throughout higher education. And then I think it also requires instructors to bring AI into their coursework, so students are actually using that in their day-to-day. And that then means that, you know, higher education should be working much more closely with industry.

So like if you're gonna be an economist, for example, how do you use AI in terms of data analytics, for example? Like, that's a very different skill set or approach, you know, if you're gonna be in allied health or nursing. And so each of these disciplines is changing in their own unique way, and we need to make sure that the teaching and learning are preparing people for that.

[00:59:13] Alex Sarlin: That gap between the 92% using the commercial off-the-shelf tools and then the 14% who feel like they're being actually trained to use it in an educational context, in an educational environment where they're actually using it to learn how to learn to build their skills rather than just to get answers or to speed up their workflows or to create new things.

I feel like that tells so much of the story of where we're at in 2026, and it really-- it's on, I think, the, the education technology and the educator, b- it's a lot of work for us to do as an educational technology ecosystem to do exactly you say, instructional training for instructors, to provide tools, to provide that full wraparound services so that instead of using these off-the-shelf tools, using Claude, using ChatGPT, using Google Gemini just in and of itself and not seeing any connection to the educational context, just sort of using it to supplement, actually having these things weaved together into coherent story.

I'm curious how Pearson is approaching this problem. As a huge leader in the field, what kinds of things are you doing to build, uh, capacity for instructors and to build tools that actually make AI work in an educational context rather than a just productivity context? 

[01:00:26] Tom ap Simon: Yes. I mean, I think firstly, I think we sort of need to acknowledge the problem a little bit, right?

And that is that AI in and of itself makes everyone's jobs harder because, you know, it becomes more easy for some students to cheat. It makes some-- more easy for some students therefore to worry about their, their friends cheating. It's-- it becomes an easy bypass. It becomes an easy shortcut, right? So we know all that as a case in the advent-- you know, in the pre-AI era, it wasn't like education was hitting the ball out of the park on a number of these fronts, right?

It was hard. It was difficult. And so in many ways, it's made the challenge much harder. And so we think that actually this requires very, very deliberate design into what we're doing. And I think when we see that done well, it makes a huge difference in terms of learning outcomes because it drives better understanding, it drives better thinking.

And we've got clear evidence that when students are asked to sort of explain their reasoning, make connections, and then actually test their understanding, that's actually all core to critical thinking. And so if you're doing it properly, then you're getting real improvements and real tangible results from a learning perspective.

And so I hate it when sort of there are these just gross generalities about AI, because that really just totally misses the point. You actually have to get really into the specifics to understand in each use case how and where AI is actually helping students learn or instructors teach. 

[01:01:44] Alex Sarlin: The critical thinking piece, I have heard that a lot recently, and it's really interesting.

I feel like there's an increasing focus on if AI can speed up some of the execution of work, it can do the algorithms, it can do the-- it can answer the factual questions, it can sort of bring context, it can obviously write essays or write papers, do research, then what the human element really becomes is, is judgment, discernment, critical thinking.

This has really been a theme recently, and I'm curious, how does that actually work in practice? You know, it makes total sense, but when you're actually designing a product, as you say, being really careful with the instructional design so that the critical thinking piece is left on the student to actually make sense of the material, to actually judge what they know, be metacognitive about what they're learning and what they need to know and what they want to accomplish, and all of these really important skills.

I'm curious if you have any principles that are sort of starting to emerge within the Pearson ecosystem about how to make sure that AI doesn't replace that thinking and instead leaves the hard productive struggle with the students. 

[01:02:45] Tom ap Simon: Yeah, totally. I mean, our fundamental premise is that actually AI, when it's built into learning in a way that increases engagement with material, that should benefit students.

So we view this as AI as a tool to help students, help them work through ideas, stay active in the learning process, and spend more time actually making sense of what you're studying. And so what I think we see is that when that's done well, we see much greater evidence of deeper thinking. For example, one in three student, um, AI inputs have been demonstrating higher order thinking according to Bloom's Taxonomy, with students then applying, analyzing, and evaluating content rather than just generating quick answers.

So sort of active reading is a huge driver, so therefore, we're making sure that we're teeing up students to be better at that. From a proficiency perspective, one of our products is called StudyPrep, and that has, um, AI-based adaptive learning And they're 90% more likely to reach initial mastery in that topic.

And that's a great example of where we're working with instructor goals to personalize learning and enhance studying. So from our point of view, it needs to be sort of tailored built in, and if it's done well deliberately, um, in partnership with an instructor, it clearly is gonna be very, very successful.

So to me, it's a lot less about the, the hype of the technology versus the intentionality of the learning design, and then actually the specifics from an implementation perspective. And, you know, many of our sales teams, when they're out talking to instructors, sometimes they have aha moments when an instructor, for example, a story I heard recently, an instructor whose mother tongue was Ethiopian, um, because we could do the translation into Ethiopian.

That's a lovely example of how the AI made our capabilities really made that instructor feel very special in terms of the teaching and learning experience. Now, that's not something that, you know, that didn't improve anyone's educational outcomes, but it did make an instructor feel more equipped to teach and more of, you know, who she was as a human being.

And that's a lovely little example of the human sort of power that this technology can give when done well 

[01:04:47] Alex Sarlin: Terrific outcomes that you're seeing in, in these studies, and I think that's a great example of the sort of personalization that can happen on the instructor side as well as the student side in terms of translation or interest-based personalization.

You can do a lot of things to make learning more accessible and relevant and, and engaging. I think the Bloom's Taxonomy is a really interesting way to look at how AI is working, and I know that some of the papers that have been coming out about it really focus on that structure because, you know, it's called generative AI, this whole era of AI, because it can create.

So technically, it can sort of reach the higher ends of the pyramid, and it can synthesize, and it can evaluate. It can do many of the things on the taxonomy. But we also know that what we want to be doing, the higher order thinking that you want students to be doing, is also at the top of the taxonomy. So ideally, AI can sort of push students up, and that's exactly what you're talking about, push them to higher orders, to applying, to synthesizing, to evaluating.

I think that's a real-- You know, Bloom's Taxonomy has its huge fans. It has some critics, but at the same time, I think it's a really, really, really useful structure for thinking about what part of the thinking process we want AI to support with and what part we want to retain as humans. You also do career readiness, and I mentioned in the, in the intro, career readiness is such a hot topic right now.

You have people making these massive investments in AI career readiness, huge changes into what entry-level skills might look like, and there's a lot of hand-wringing and a lot of guesswork. Pearson is really, really a large career readiness engine. You do a lot of different products and a lot of different thinking about helping especially higher education graduates move into their workforce.

How are you thinking about career readiness at this extremely complex moment for the workplace? 

[01:06:28] Tom ap Simon: That's a great question, and I will never do it justice on this call. But, you know, I would say a few thoughts in, in no particular order. I think firstly, career readiness starts in middle school And so if you stop and take a step back, what's been true for a number of years is work, higher education, and K-12 have been blurring, and that's increasingly happening at a, a much more junior level now.

So, you know, one of the themes that we've all seen, right, is the, the, the much greater sort of penetration of dual enrollment in community colleges, uh, which is great in terms of, like, a- advancing ROI and, and getting kids more set up earlier in their academic careers in terms of college credit, confidence, you know, cheaper cost and so on and so forth.

I think all of that's just been sped up significantly because actually what you're seeing is kids really-- or parents waking up to traditional pathways from an education perspective no longer necessarily being right, given the whole debate about ROI in, in higher education, and clearly it works for some students, and clearly it doesn't for others.

So I think that there's a much greater appreciation that it is not a one-size-fits-all path, but actually, you know, you can get a really good exposure to the skilled trades, to different avenues, different pathways much earlier in school, and then that then helps you understand as you're getting more sort of streetwise as you're getting into high school what direction you may want to go in.

So, you know, we're really seeing, you know, a huge opportunity in middle school and high school here, and then also in, you know, higher education as well. So, so we view it as it's not, "This is the right path for Alex, and this is the right path for Tom." You know, if you do a bunch of CTE courses at high school, you're more likely to be more successful in a two-year or four-year setting, and you're more likely to graduate high school.

So it's not like we have to have this track or we have to have that track. There are lots of different tracks, and I think we just gotta get all much more comfortable with, with acknowledging that and recognizing that and being really intentional about supporting each student where they are in, in terms of where they want to go.

Now, the key word for me in all of this is applied Whether it's whether you're talking about AI, whether you're talking about, you know, simulations, whether you're talking about working as a med tech or an allied health or in HVAC, for example, the learning has got to be relevant for the career. Because otherwise, understandably, industry's then gonna be jumping up and down going, "What am I meant to do here?"

And so that I think is the real thing that we've got to get right, whether it's AI or skill-based learning, it's got to actually be really, really relevant for the student, where they are and where they want to go. And that's easier said than done because, of course, there's so much change happening in the labor force right now, it's really hard to get really good, easy signal in terms of, "Ah, well, we should go off that."

You know, because of course, what we're seeing is there's so much fracturing in terms of new jobs, how they're defined. You and I might view, you know, a new job very differently, and we might categorize it very differently. That therefore makes it hard to get reliable labor market signal. And so we're doing a lot of work in terms of getting really tight in terms of what those pathways look like and, and what the new opportunities are in the future.

And that, by definition, requires us to work more closely with industry partners because if we do that well, then we're more in tune to them and, and we're better able to think through what qualifications we can help support them with. And so that's some of the f- the work we're doing. And we have this, you know, this fantastic business that we have called Certiport, which does a whole host of Microsoft Office certifications, Adobe certifications.

And I was talking to somebody on Monday, and they were explaining how the school administrator needed some help with Excel, that they sent in one of their high school kids who was, you know, exceptionally good at using Excel because they'd done a Microsoft Excel course, and they're competing in our national championships today on it.

And they went, and they did it all in five minutes, and the school administrators had no idea what the, the kids had actually done because they'd been doing some advanced Excel. But, you know, imagine the confidence of, of a 17-year-old kid if you can actually go in and help in s- the high school administration and actually make a real difference.

And so that's, I think, when done well, is a beautiful example of real skills working really well in really tangible environments that make, um, meaningful strides forwards in people's confidence to do whatever they need to do. But, you know, like if we can get kids set up feeling confident to approach the world, that's half the battle done in terms of actually making sure that they're both confident and they understand what they need to do.

[01:10:39] Alex Sarlin: Absolutely. And I'm hearing some trends you're mentioning that I think you build together into a much more comprehensive and sort of cohesive perspective than I've heard from a lot of people about how the career pathways will work. I mean, there's pathways coming in earlier and earlier to higher ed and even to high school and middle school.

The idea of skilling becoming relevance and relevant skills that are career aligned coming in earlier and earlier into education, but also not trapping people into a specific track, especially as we are so aware how much things are gonna change. You know, it strikes me that, that AI is-- You know, at this moment, you know, we think of AI sort of as a almost like a career pathway or it's, it's a-- I don't even know how people think about it, but because we're so early in the AI era, it feels like the first few years of the internet era where we're like, "Are you gonna have an internet job?"

And it's like, that didn't make sense anymore, right? You don't have an internet job anymore. Every job involves online work. It involves computer work. AI is the same way, you know. Every job is going to involve it. So I'm curious, when you talk about allied health or pathways for all the HVAC, all these different types of jobs, are you thinking about weaving AI into the pathways so that just expecting that every industry is going to be influenced by it in the future?

[01:11:49] Tom ap Simon: Absolutely, right. I think that is the most natural way of thinking about this, because of course, AI is gonna change, you know, how we teach the fundamentals of any of those disciplines in the first place, just as it should, just as it-- the com-- the calculator did, for example. That is A-okay. And so then the question becomes a little bit: How do you get faculty and students set up to learn?

Because each of those fields is going to be influenced in their own way, right? And so, you know, how you become a med tech is gonna change, how you become a nurse is gonna change, how you think about HVAC is gonna change. And so fundamentally, you have to weave AI into the curriculum and the learning specifically with as many applied examples as possible, because without application to the real world, like if I sit here and I go, "Oh, well, you know, I'll come up with something in an ivory tower," but that's not helpful to students, right?

But if I sit down and I work with an industry partner, and then I'm like, "You know, how would this work well for you in terms of thinking through some simulations or some real use cases that your earlier career employees are gonna bump into?" Then that makes it much more tangible. The employer feels much more confident that it's actually being used properly.

And so it-- to me, it's clearly going in that direction. I would also say it's about rethinking how we do teaching, because clearly we've had a model where it's been, been based around, for example, points for homework, right? That was obviously a model that worked perfectly well when students were incentivized to get points for homework.

But we know there are so many different models out there or temptations now, not necessarily to do the hard work. We need to reinforce productive struggle, right? You and I left to our own devices, we're both gonna sit there, probably prevaricate for a little bit before we knuckle down and, you know, really focus on the learning.

That's what human nature is about. But learning is hard Productive struggle is hard. It's difficult. And so we shouldn't shy away from that. We should be very upfront about it. We should be very intentional about what it looks like, and we should be very clear in incorporating AI and applied, uh, you know, applied learning examples into it, so we actually get students set up to be successful.

All of that sounds really easy when I say it in thirty seconds on this show. It's really, really hard to do in practice, and it requires a lot of time and effort and thought and deliberation to get it right. And so I think that's what's really interesting right now because I think that there are lots of loops that are sort of building on one another in terms of understanding the data, understanding the learning outcomes, understanding, you know, if you've got a particular graph and, you know, sixty-four thousand people tried to answer a question on the graph and got that slope right or wrong, what do you then need to do in terms of that next step for them from a learning perspective?

Again, it's gotta be real, it's gotta be applied, and it's gotta be grounded in where they're going. But I think that's some of the things that we're potentially seeing at the, the sort of the hype cycle, which could potentially be real over the next couple 

[01:14:33] Alex Sarlin: of years. Absolutely. And I think your focus on what you're mentioning about sort of the points-based homework paradigm of right, th-that you're going home and you're trying to do something and get it correct as fast as possible and sort of work through the problem set is really such a different way of thinking than the AI era, where you're actually trying to sort of decide what to do, focus a direction, discern, use some judgment, struggle with it, not just try to get the answer as fast as possible, because that's exactly what AI is good at.

It's actually think about the problem, think about what you want to get out of it, think about your own learning and how to make it sink in and actually wrestle with something. And it's like it's a real paradigm shift to think about homework or to think about schoolwork in general, rather than sort of jumping through hoops as giving more autonomy and sort of power to the students to actually use their own thinking and think about what they want to accomplish.

But it's also very exciting to hear your point about the student with Excel. It builds confidence. It builds empowerment if you feel-- if students can feel like they can really make their own choices in their education. I, I look forward to that type of education. I wish we had more time, but we're coming right at the end of our window today.

Thank you so much. This was a really interesting conversation. Tom ap Simon is President of Pearson Higher Education and Virtual Learning. He's leading teams that serve millions of higher education and K-12 learners globally. Pearson absolutely, you know, pioneer for a hundred plus years in the space. Thank you so much for being here with us on EdTech Insiders.

[01:16:00] Tom ap Simon: Alex, thank you so much. I've had a great time. 

[01:16:03] Alex Sarlin: Welcome to the EdTech Insiders Podcast. We have a special guest this week. We are talking to France Hoang, who's the founder and CEO of BoodleBox, which is a collaborative AI workspace. We'll find out exactly what that means. It's really interesting. He previously helped build companies with over $600 million in sales across tech, defense, and law, and brings leadership experience from the White House, from the US Army Special Forces, and from federal law.

A little bit of an unusual background for an EdTech founder. We're really happy to have you, France Hoang. Welcome to EdTech Insiders. 

[01:16:39] France Hoang: Alex, it's great to be here. It's fantastic to join you. Just so people wonder, like, what's a person like me doing in education, I also have a deep love of education. I've served as a university trustee, and I, I became and remain a, a distinguished visiting lecturer at West Point.

So I've done everything from, "Open your textbooks to page 46," to, "Who should we pick as the next provost?" 

[01:16:59] Alex Sarlin: Fantastic, and, and, and that's a good thing, just to be clear. I think people with different experience in the White House, US Army Special Forces- Yeah ... have incredible experience to bring to EdTech. I'm very excited about that.

Yeah. So let's talk about what BoodleBox is doing for higher ed. You know, you, you've argued that higher education is evaluating AI the wrong way, and we're definitely in a, a morass when it comes to figuring out what to do with AI in education. What do you think colleges and universities have been missing about AI, and how is BoodleBox stepping in to really clarify the picture?

[01:17:31] France Hoang: Yeah, Alex, let's, let's kind of back out 50,000 foot, right, and just talk about work in general. Because, you know, obviously, what we do in education has to tie to the workforce in some way. We are, we are preparing tomorrow's workers today in our classroom. And for most of human history, work looked one way, right?

It had one shape. We spent a little bit of time defining the problem. We spent a lot of time generating a solution, designing, creating that solution, and then we, we validated that solution really quickly, and we spot checked it. And generally speaking, colleges a- and universities prepared students for the middle two phases, being great people at taking somebody else's direction of what a problem is, creating a solution, then somebody else would check the work.

And so work had a certain curve that looked like a mountain, right? A hill. Well, AI has inverted that. So if you think about it, in an, in an AI era, we now spend the bulk of our time actually defining, identifying the problem, giving the AI context And then the AI does the vast bulk of the work designing and creating.

And then because humans weren't involved in designing and creating, we just spend a lot more time validating what AI created. And so now the work has become, instead of looking like a mountain, it looks like a valley. And I call this the Lorantz curve. Chris Lorantz at West Point introduced this to me.

And those, the last step and the first step, the, the define phase and the validate phase are what I call the discernment phase. They are what requires human discernment and judgment, whereas the, the b- the middle two phases are what I call the generation phase. School has mostly been about generation.

There are folks now who argue that we should not generate with AI at all, right? And that we should continue, we should reject AI and continue to do work the old-fashioned way. There are other folks who say, "No, we should be embracing AI and using AI to generate faster and better than ever before." I would argue to a certain extent they both miss the point because they're still focused on generation.

Mm-hmm. Whereas I think the human value is now, and maybe arguably always has been, discernment, right? What is the problem? Do, do our solutions work in the real world? And so I argue that when we think about AI in education, we should be focusing on creating human actual intelligence, wrapping generation with human discernment on both ends of it.

And so BoodleBox is a collaborative AI platform to teach with and about AI to build that actual intelligence, to create the abil- the human ability to discern while also generating with 

[01:20:08] Alex Sarlin: AI. Hmm. So let's dig into that. The ability to discern, the ability to set a path, to set a goal, to figure out how- Mm-hmm

you want the world to change or processes to change or improve society or improve a certain type of, uh, outcome. I, it's a really interesting take that discernment is now sort of the primary task of humans- Mm-hmm ... in an age where AI can handle so much of the productivity and generation. How does collaboration play into this, and what does it look like in practice in a higher education platform if you have people collaborating- Yeah

to discern? Walk us through it. 

[01:20:41] France Hoang: Yeah. So if you think about problem identification, right? What does it take to properly identify a problem, right? It means you know what questions to ask, and it means you know what context you're providing. It means you know what, how to evaluate what comes out of solutions.

And previously, that required you to, um, have a lot of life experience, right? So typically speaking, we train people to be analysts or junior associates or, or beginning knowledge workers, and all you were expected to do was to be able to generate what somebody else told you to do. You show up your first day of work and your boss says, "Make this spreadsheet for me."

You make the spreadsheet, and the boss checks it over. Well, in an AI era- Right ... we now have to be the ones who figure out, because the AI can do the spreadsheet, we have to do what our bosses did, right? We, we can't rely on years to get promoted to b- earn the right to be a discern- you know, discerner. We have to be a discerner right away.

So I would argue education now needs to teach discernment from day one. And so how do we do that? One, you know, one of my favorite quotes about education is from Plutarch. He said, "The mind is not a vessel to be filled-" but a fire to be ignited. And so, you know, I hate to say it, but I think in many cases, education has been just about filling minds, right?

Here's knowledge. You know, put this knowledge in your brain. Well, knowledge is now a commodity, so we don't need domain knowledge, we need domain expertise, right? That not-- That's not just knowing things, but knowing what questions to ask about those things, knowing about the context. And so education, first and foremost, should be about creating that domain expertise.

Second, it is important to know how to use AI, right? To be a good generator, 'cause those are the tools that are out there. And so knowing when and how to use AI responsibly, even if at all, right? And knowing when not to use AI is, is an important skill set. And, you know, AI literacy is still very, very important, right?

But third, I would argue the things that matter most now are the things that make us uniquely human, right? Critical thinking, collaboration, communication, uh, creativity. And in an AI era, that means Being able to collaborate in human AI teams. And so what does that look like in a classroom? It's a classroom where you're using AI to create domain expertise, right?

Using teaching in ways you couldn't before. Personalized instruction, AI tutoring, simulations. You're simultaneously you're teaching about AI to give that AI enablement, right? And not just learning a tool, not just ChatGPT or Claude or Gemini, but learning about the technology, because as we know, these tools literally change week over week.

Yes. And so you don't want to just master a particular tool, you want to understand this technology so that no matter what tool is the flavor of the month, you're like, "I'm-- I got this," right? So, and then third, we don't know how to operate in human AI teams. This is like a brand new skill set, right? We have-- Think about it.

We, we send people to MBA programs or military academies or leadership programs to learn how to manage people. Well, now we have to manage people and AI. Yeah. This is like complete greenfield. And so I would argue the place to start is just having a space where you can have multiple people interact with multiple AIs just to, just as a, just as kind of step one of that, and BuduBox enables that.

So inside BuduBox, you can literally have a team of students working with a team of AIs together and do it in a way that is transparent to the instructor. And I'm, and happy to provide an example of that here in a second. 

[01:24:02] Alex Sarlin: Yeah. I, I would-- I'd love that, and I'm sure our listeners would as well. I, I-- and I'd love to hear the combination of the, the collaboration between humans and how that leads to collaboration between humans and AI, because it feels like that, as you said, it's a really interesting world that we're coming into, where the first day of work you might have to actually make some, some strategic decisions and not just executing, because that execution is, is, uh, is outsourced for many, many, many different things.

So what does it look like in-- Yeah, give us an example. What does it look like with students to be working together and then managing a team of AI agents or working with AI in various ways to accomplish something that they've sort of come up with themselves? 

[01:24:40] France Hoang: Yeah. So I use this example a lot. Um, Professor Christina Kallberg at, uh, Point Loma University teaches marketing.

She does, you know, the typical sage on a stage lectures in front of her class. But after that Her class becomes really interesting. So her students, after their classes, are required to take their lecture notes and then build an AI assistant, build their own marketing assistant that is trained on their lecture notes.

Hmm. And then her assessments and her assignments are not write an essay or do a quiz. They are, here's a real world problem. I want you to collaborate with your AI assistant and generate a real world solution. 

[01:25:20] Alex Sarlin: Mm-hmm. 

[01:25:21] France Hoang: And then as her class goes along, they have to do that in teams of students and teams of AI assistants that the students created, and then they bring in real world, right, actual clients with real world problems.

So by the time the student completes her class, they've learned marketing, they've learned how to build an AI marketing assistant, and they've learned how to collaborate with those AI marketing assistants One-on-one and then collaborate in teams of people with their own AI market assistance. So think about how better prepared that student is than someone who attends a traditional marketing class, and also, right, that's someone who can immediately discern, right?

They, they are-- been working on not just, "Oh, t- someone tell me what to do. Okay, here's a solution." Here's, here's a situation. I need to identify the problem, I need to use AI to then design and create a solution, and I need to go back and validate that this AI-generated solution is actually valuable and not just AI work slop that looks good, right, but actually isn't good.

And I think we've all had the experience of seeing how spectacularly yet confidently wrong AI can be. 

[01:26:28] Alex Sarlin: Yes. Or people using AI can be. 

[01:26:32] France Hoang: Absolutely, Alex. 

[01:26:33] Alex Sarlin: Yeah. Yeah. You're bringing up some really interesting points, and I wanna double-click on them because I think it's really-- it, it's a key moment right now where AI is so-- You mentioned it's sort of such a greenfield set of skills to work alongside AI.

A- and you're right, I mean, just nobody has their head around it yet. It is so new how to even do that. And you see things online where people are like, "I built a thousand AI agents and they're doing all my work for me." And somebody else is like, "Oh, I built my go-to-market team and it's all AI." And you know, y- there's so little validation of, like you say, what is AI work slop and what is effective.

And I'd imagine that within a real-world context, what's effective is what's working, right? If you built your AI go-to-market team and it's successfully go-to-market and you're, you're making it work, then hey, something worked. But in a classroom environment, it's sometimes less clear because it's not always associated with real-world outcomes, and I know this is a passion of yours.

Tell us a little bit about what it looks like w- within the BoodleBox world to try out something in a classroom environment and then actually see if it works in a, in a, in an authentic environment. 

[01:27:38] France Hoang: I love that question, Alex. So first I wanna draw a contrast. So if you think about native AI apps, right, whatever, pick, pick the app of your choice, they are all single-player tools designed for individual productivity, right?

The whole point is for you or me to use a particular tool and get something done, and the-- they want it as frictionless as possible. But education isn't an individual activity, it's a collaborative activity. 

[01:28:05] Alex Sarlin: It's true. 

[01:28:05] France Hoang: And the point isn't productivity, the point is actually product- productive struggle. In other words- Mm

we want friction as an educator, right? I h- I have to have friction as a student. To, so to that-- If you look at it through that lens, the native AI applications are actually the absolutely wrong mode for learning, right? Because they wanna make things as easy as possible. So BoodleBox makes it possible to reintroduce friction and productive struggle.

And so in BoodleBox you have what's called the AI classroom, and it-- Within an AI classroom I can create an assignment and say, let's say, Alex, I'm trying to teach you about, let's say I'm trying to teach you about the conference at Yalta, right? This historical event at the end of World War II. Yeah. Well, if I just gave you a reading assignment and go, "Hey, go read these 60 pages," you're just gonna stick it into ChatGPT or Claude or Gemini, get a summary, show up to class, and then wait for me to call on you, and then you're just gonna look down and read a couple of bullet points, right?

No friction, because AI's made it frictionless. What if instead, right, I create an AI simulation inside BoodleBox, and your assignment isn't to read, your assignment is to actually be a part of Yalta. Take on the role of Stalin or Churchill, um, or Roosevelt, and actually go through the conference and then make decisions, and then get feedback on your decisions and what happens, and learn about the actual conference, but then engage in critical thinking and develop discernment.

And then when you get to class, we're gonna have a discussion about what decisions different students made- Mm ... and reflect on that and what they learned from that. And so we've introduced a, not just struggle, but a different level of struggle, because the struggle isn't in generation anymore, the struggle is in discernment.

Mm. And so every possible opportunity I have, I don't want you just to be better at having knowledge and regurgitating the knowledge. I want you to get better at, at discernment, identifying what are the problems here, right? And then validating whether the solutions that are being offered to you are the right solutions.

[01:30:03] Alex Sarlin: Yeah, that's very interesting. And I, I, it strikes me as an instructional designer that, you know, that type of simulation, what you're really asking people to do is, is, is apply the knowledge, is go much more deep into actually using it. If you're Stalin at Yalta, you need to know what Stalin wanted at Yalta.

What does Russia want at Yalta? Mm-hmm. What does the USSR want at Yalta? Who-- What does anybody else want? How are you negotiating? It's a whole different way of understanding the situation because you're actually thrust into it than understanding the facts and the sort of just core knowledge behind it, which then, which can be sort of chewed out and regurgitated by AI.

I think that's very powerful. And, and so-- and what role does the, the, uh, educator play in that type of thing? Is the educator making these simulations? Are they evaluating the outcomes of them? Yeah. Are they-- How, how does it, uh, sort of- Yeah ... work within an, uh, environment where you have a human educator at the core of the experience?

[01:30:56] France Hoang: Yeah. And I love what-- I love the way you framed that, Alex. The human e-educator is absolutely the core of the experience. You know, I do a lot of, of visiting of, you know... Buddle Box now was just told it's now working with a hundred and sixty-seven colleges and universities. You have over a hundred thousand faculty and staff that use Buddle Box.

And I, I often get this wonderful opportunity to go to different campuses and talk to different instructors. And there's always- Yeah ... this u-undercurrent of concern, like, are, is higher education relevant still, right? Are, are faculty and educators gonna be replaced? And I don't mean to laugh, but, like, farthest from the truth.

I actually think educators are more important than ever because, again, AI can do all the, can do as much design and creation as you want, but the discernment, that's a uniquely human activity, and it requires ex-lived experience. It requires an understanding of pedagogy, right? It under- requires expertise.

And so, you know, the instructor is, is more important than ever because teaching isn't a series of tasks, right? It's a, it's a lived experience with your students. And that human connection, um, you know, I think one of the promises of AI is the teaching profession, just like any other knowledge worker, I think up to this point has consisted of eighty percent drudgery and twenty percent of the things I actually love, right?

Connecting with other people and being creative. One of the promises of AI is it can flip that. What, what if AI allows us to do all the drudgery parts in twenty percent of our time, we can spend eighty percent on the great parts So an educator in BuduBox is the orchestrator of all this, right? They get to decide what the learning objectives are, and then they get to decide what form their assignments take.

Because BuduBox enables evaluation of not just the product of using AI, but also the process and seeing the progress a student makes, th- there is this rich vein now of, of potential feedback and assessment that wasn't possible before, and it's designed to be interactive. So, um, one of the things we built in BuduBox, it's a, it's a beta feature right now.

Um, it's called Insights. I jokingly refer to it as TA Bot. It's an AI agent that has access to everything you have access to as an instructor. So imagine an AI that can literally see every student assignment, every learning objective, every reading, and the things you can ask it to do, right? I mean, on a super granular basis.

And so there's a lot of frankly, new potential as a faculty member, right, as an educator to, to transform your classroom and to advance your, your students' learning in an era of AI 

[01:33:38] Alex Sarlin: Yeah. So I guess the, the follow-up question to that is, in an era where everybody is new to this world of, of AI and human interaction, and AI doing so much of the execution and humans doing the discernment, this, this framework you're laying out, which I think is really powerful, how do educators sort of flip that script in their own head of, of, you know, "Hey, I'm used to teaching Yalta and making sure everybody knows, you know, who was there and what they wanted and what-- when it happened and why it was important, and the context historically," to being like, "Well, now I can create simulations around it.

I can create deeper learning objectives. I don't have to worry as much about the, the factual underpinning. I mean, I w- I want students to know it, but I want them to go deeper and get really involved in it, do a simulation." How do you help instructors make that leap to realizing what's possible with AI and how to build a, a more, you know, richer and, and more in-depth learning experiences, more active learning experiences?

[01:34:38] France Hoang: Yeah. Look, I get it. It's a little terrifying, right? Like, 

[01:34:42] Alex Sarlin: I, I 

[01:34:45] France Hoang: I, I'm used to-- You know, I think those of us who, who teach, and as I mentioned, I'm a visiting lecturer at West Point, right? We-- Th- there's something powerful about being the sage on the stage and standing up. Right. No matter what question a student answers, you like, you're dropping wisdom left and right, and it's like you're a master of your domain, right?

I think the first thing we have to acknowledge is that in an era of AI, that's just not gonna be true when it comes to AI. Like, you have to give yourself permission to learn. You don't have to master this before you start teaching this. In fact, I think some of the most powerful examples of, of AI in a classroom I've seen are professors who do the opposite.

Beginning of the semester, they stand in front of their students and say, "Look, I'm gonna use AI in this classroom. I'm gonna use it for my own lesson preparation. I may use it for some assessment. Um, I'm gonna try a bunch of things. I'm gonna try them with you. I'm gonna share what I'm doing. I'm probably gonna make some mistakes.

You're probably gonna tell me some of these things aren't effective, and that's okay 'cause I expect you to use AI and experiment and share what you're doing, and we're on this learning journey together." Yeah. And giving yourself permission to learn alongside your students when it comes to AI. Now, with regards to how we actually use it in the classroom, again, this is great, Bill and Alex.

Like, we've never had such a, a transformative technology introduced, right? And be- and to be clear, those of us, you know, in education, we did not introduce AI into our classrooms. Let's just be clear about that. It was introduced for us Right? And so we're, we're gonna have to gr- we're grappling with this in real time.

So we understand that. So inside BudleBox there's a number of tools and things- Yeah ... that we've done to make this transition easier. First, there's a bunch of toolkits for, for faculty and for students. Uh, one of the challenges of AI is because it can do everything, you don't know where to start, right?

Right. You don't-- Just too much. And so these toolkits identify specific use cases, you know, generate a syllabus, um, create a rubric, um, prepare for a class lecture, whatever, that are very, very granular and give you an idea of the use cases of AI. Second, we've built an AI coach So it turns out AI is great at coaching you on using AI.

Yeah. And that coach, if you turn on memory mode, actually remembers you from chat to chat and from session to session. And so it's an AI coach that is coaching you over time to improve your use of AI, and it understands you and it understands your role as an educator, uh, which is fantastic. We also believe in the power of community.

Um, I often get asked, "What's the most powerful thing we've built at Budobox?" My answer is always the same: it's our community. We run a cohort program. A hundred educators join every month and get to connect with other educators using Budobox and exploring AI, really powerful peer-to-peer, uh, learning. And at the institutional level, we have a number of programs where we, you know, we go on site, w- you know, we help schools plan their implementation of AI, and we actually have a number of, of experts that if a school wants, can come in and, and assist with that process.

[01:37:43] Alex Sarlin: That's-- Yeah, that's a powerful set of tools. So there, there's templated toolkits to help people make sense of all the different possibilities. As, as you say, it can do anything, so where to start is-- it can be really tricky. I feel like that's true a-across the board for people in, in every area of education and beyond.

And you have AI coaching itself, so that sort of meta AI can help you if you say, "Oh, I'm teaching Yalta. What might be some interesting ways to, to use simulation and/or to, to, to make it more engaging or to raise the, the active learning?" And then you have, uh, this community of humans working together to try to experiment.

Uh, I, I like your example of the teacher, of the professor or teacher who stands in front and says, "This is new. I'm, I'm gonna, you know, I'm gonna be making mistakes. I'm gonna be trying new things, and I expect you to be making mistakes and trying new things, and we're all gonna be working with this together."

I, I think- Yeah ... you know, we're, we're at this interesting moment right now where I think some people are going increasingly that way. They're saying, you know, "This is obviously incredibly powerful technology, and I have to wrestle with it to figure out how to use it in the way that, you know, how it's gonna make my life better and make the lives of my students better."

And other people saying, "This is incredibly powerful, p-p-powerful technology. I don't know where to start. I know that there's-- it, it, it empowers everybody to do all sorts of things they could never do before, and I don't know what that means for my wo- this world as a, as a professor or as a, a h- a student in, in higher education."

I'm curious, you obviously have a really well-thought-out philosophy about way-- how AI can help, how AI can create discernment, how AI can work with humans. What do you say when people come in the higher education context and they're skeptical or concerned, or they've tried this in the classroom and it hasn't worked the way they wanted?

How do you sort of get them back to a place of, of being open to trying it again or open to diving in? It's something that I think a lot of people are wrestling with right now in the space. 

[01:39:35] France Hoang: Yeah. So first off, I didn't s- start and build Buduax 'cause I love AI. I frankly started and built this company 'cause I'm concerned, deeply concerned about AI.

So I think their skepticism is well-founded. Every time we have a new technology, we have to wrestle with it. Yeah. I am concerned about getting the balance and the relationship between humans and AI right. You know, I think there's a good chance- Yeah ... we might not get it right if we don't do the right things, right?

And so I wanna be on the good side of that equation. Second, I often get asked, uh, when I do keynotes, almost invariably somebody in the audience asks me questions like, "What do I do about my faculty member who refuses to use AI?" Exactly. Um, my answer's always the same. Yeah, my answer's always the same. You don't do anything.

Um, we value academic freedom. We value academic diversity. I think it's good for students to have-- to be able to choose between a classroom that uses AI and one that doesn't. I think professors should have the right to t- to teach that way. Um, I think there is a value in having d- you know, there's different styles of learning for different styles of learners, and, uh, faculty should cater to that.

And I think that's perfectly fine. I don't think everyone needs to use AI. And to your point earlier about AI use in the corporate world, I think in some cases we are using AI because we can, but it may not make sense, right? Just because you can doesn't mean we should. Um- Right ... I see lots of examples, even on LinkedIn, right, of people doing the equivalent of taking a Formula 1 racing car to the grocery store, right?

Right. And it just thinks they're like, "Look at this cool thing I did," right? But is that really necessary, right? Like- Right ... is it the best way, or is it just the cool way to do it? And so again, AI, I am by no means saying, like, all AI, all the time, everywhere, anyhow, anytime, right? Like, I think there's plenty of academic use cases where we should go back to the blue book or an oral examination or teaching without AI.

I think it's a tool, right? I think it's an important tool. It's a transformative tool. I think there are-- We certainly need-- We can, we can teach with it, we certainly should teach about it, um, but it's not always the answer. 

[01:41:39] Alex Sarlin: Yeah, I think that's a really powerful answer . And, and you know, I-- the, the fact that you're building an environment with BoodleBox that allows collaboration, allows human discernment, it allows humans and AI to work together or to collaborate amongst themselves, you're obviously thinking really deeply about creating really rich learning environments that are teacher-centered, that are student-centered, that are really, really thoughtful.

And yet at the same time, I love that answer. You know, what do you do about the faculty member who's not using AI? You don't do anything. They're not gonna use AI and that gives everybody, you know, options. I think that's a really powerful answer, and one I've actually never heard before o-on this show, so I, I love hearing that.

I wanna ask, you know, you do-- Just to circle back to the beginning of the conversation, you do have a really, uh, interesting background. You've worked in, in the Whi-White House, you've worked in the US Special Forces, you've taught at West Point. You still teach at West Point. One thing that's always struck me as very intriguing about West Point specifically, or about military education, is that there's a case to be made that military education should be the most practical education, right?

If people are training to go into the military, they're-- they, they need to be able to do very specific things. And it's, it's in-- You could consider it vocational training, but it is not at all vocational training. West Point is incredibly academically rigorous. It teaches all sorts of philosophy, it teaches all sorts of history, and I've always found that a really interesting approach.

And I'm curious how that sort of approach of combining practical education with, you know, traditional classical education has informed your teaching style and informed your work with BoodleBox 

[01:43:07] France Hoang: Yeah. One of the, one of the most apt descriptions of West Point, and, and they talk about this in- inside West Point, is Athens versus Sparta, right?

Hmm. To what extent is, is West Point about Sparta and the mili- it's the military academy and training, very practical, right? You, you're training young men and women to be managers of violence. If you think about it, right, at the end of the day, that is what an army officer has to do, right? Yeah. At the same time, we want to be Athens, right?

We wanna be, we wanna be academics and scholars, and one of the most fascinating classes I took at West Point was, um, phil- introduction to philosophy, which included multiple sections on jus ad bellum and jus in bello, right? The study of, of the just- justness of war and in war, and we had debates, and we even asked questions like, "Should we have dropped the second bomb in, in World War II on- Yeah

on Nagasaki?" Right? These, uh... Sometimes I tell this to people and they're surprised. They think that, like, West Point's just training you to be this, I don't know, this, this automaton, right, who follows orders. Yeah. And it's far, farthest, farthest from the truth, right? It's actually the exact opposite. We want thoughtful- Yeah

leaders of the military. And so I think, I, you know, I'm a big fan of, obviously of education. I'm a big fan of applied education, and I think there's a role every, uh, I, I think every institution to a certain extent, even if they're not a military academy, should be both Sparta and Athens, right? There are things that we study because they have intrinsic value unto themselves, but w- ultimately, we have to apply these things in the real world.

And so I think BudBOX is d- is certainly an extension of that principle, right? There are, there are certain things that we believe in because we believe in them, right? Environmental consciousness, right? We reduce token reduction. We have, we have certain beliefs around the fact that, like, we shouldn't just be right, but, but do right, and, and those are design decisions.

But ultimately, at the end of the day, you know, we need to solve real world problems. Um, we need to make sure that students are prepared for this AI-enabled future. We n- staff and faculty have to be able to teach, right? And be able to assess that work, and so, you know, while, while you can have certain principles, they have to express themselves in ways that have an impact in the world.

[01:45:21] Alex Sarlin: That's a fantastic answer to what I thought was a curveball question, and not even close to curveball. You knocked that out of the park. Athens versus Sparta, I think gave us a lot to think about. I love that. And yeah, I, I think that that is a really fascinating model for the future of education in an AI era.

How do you combine, you know, a- academics and understanding the world and history, philosophy, humanities, making sense, you know, what do we want to happen? That to your point about discernment, what do we want the world to be like, and how might we help that world be more like that, be the change, right? With this, uh, incredible power of execution that, that is just exponential, and we're all trying to figure out what does it mean when you can make anything happen.

You can create inventions, you can create companies with one person. It is a very strange world, but I love that Athens versus Sparta metaphor as a way to make sense of it. This has been fascinating. Uh, France Hoang is founder and CEO of Boodlebox, collaborative AI workspace working in, what'd you said, 175 universities already?

Amazing. 

[01:46:23] France Hoang: Probably by the time we finish, 176.

[01:46:25] Alex Sarlin: There you go. There you go. Right. Continuously growing. Really, really, really interesting work, and I appreciate you being re- here with us to share some of your philosophy and some of what you're doing with Boodlebox on EdTech Insiders. 

[01:46:36] France Hoang: Thanks, Alex. Great to be on.

Appreciate your time. Thanks for listening to this episode of EdTech Insiders. If you 

[01:46:42] Alex Sarlin: 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 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