Edtech Insiders
Edtech Insiders
The Future of Feedback and Powering 40 Million Learners with Learnosity CEO Gavin Cooney
As CEO and Co-Founder of Learnosity, Gavin Cooney still talks about the company he helped build over a decade ago with an infectious zeal. It’s easy to see why: what began as a project in a garden shed (a cliché maybe, but 100% true!) has matured into an internationally renowned tech company with a 200-strong crew of problem-solvers. Since Gav believes that all social ills can be remedied through education, his modus operandi is pretty simple: give education the standard of technology it deserves.
💡 5 Things You’ll Learn in This Episode:
- How Learnosity empowers EdTech platforms with AI-driven assessment tools.
- The transformative role of AI in reducing teacher workloads and enhancing feedback.
- What makes Learnosity’s essay-scoring tool uniquely efficient and cost-effective.
- The importance of specialized EdTech solutions versus general AI tools.
- Gavin’s vision for the future of AI and assessment in education.
✨ Episode Highlights:
[00:02:23] Gavin Cooney shares Learnosity’s global impact: 40M users and 19B questions delivered annually.
[00:03:38] How AI-driven essay scoring saves teachers hours of grading time each week.
[00:06:48] Learnosity’s cutting-edge AI achieves unmatched scoring accuracy at just one cent per essay.
[00:10:25] Why Learnosity focuses on enhancing teacher feedback rather than replacing educators.
[00:13:54] The balance between general AI tools and Learnosity’s specialized capabilities.
[00:18:10] Learnosity’s strategic pivot to invest heavily in AI development and innovation.
[00:24:02] Gav’s optimistic outlook on AI as the “100-year wave” reshaping education.
😎 Stay updated with Edtech Insiders!
- Follow our Podcast on:
- Sign up for the Edtech Insiders newsletter.
- Follow Edtech Insiders on LinkedIn!
🎉 Presenting Sponsor:
This season of Edtech Insiders is once again brought to you by Tuck Advisors, the M&A firm for EdTech companies. Run by serial entrepreneurs with over 25 years of experience founding, investing in, and selling companies, Tuck believes you deserve M&A advisors who work as hard as you do.
[00:00:00] Gavin Cooney: You have a very finite number of engineers, so do you want to have them reinvent a wheel of doing essay scoring or whatever, or do you want to kind of leave that to the experts, especially when we're doing it, send an essay and that's kind of, it's just, it's wildly cheap, right? So do you want to do that?
Or do you want to kind of focus your developers where they'll really add value in your product and to your students? And that's the whole. The whole promise of theirosity and what we've been doing that that kind of leveraged model where we'll do something and it'll reach more millions of students and it means we can kind of we can invest more in any one thing and any one of our customers or any one publisher or any one edtech product.
Could ever do.
[00:00:45] Alex Sarlin: Welcome to ed tech insiders, the top podcast covering the education technology industry and funding rounds to impact AI developments across early childhood, K 12, higher ed, and work. You'll find it all here at ed
[00:01:00] Ben Kornell: tech insiders. Remember to subscribe to the pod, check out our newsletter and also our event calendar.
And to go deeper, check out ed tech insiders plus where you can get. Content access to our WhatsApp channel, early access to events and back channel Insights from Alex and Ben. Hope you enjoyed today's pod
today. We are joined by longtime EdTech insider's friend Gavin Cooney. CEO of Learnocity. He's also the co founder. Gav still talks about the company he helped build over a decade ago with an infectious zeal. And it's easy to see why. What began as a project in a garden shed, maybe it's a cliche, but 100 percent true, has matured into an internationally renowned tech company with a 200 strong crew of problem solvers.
Since Gav believes that all social ills can be remedied through education, his modus operandi is pretty simple. Give education the standard of technology it deserves. Enjoy our conversation with Gavin Cooney, CEO of Learnocity. Hello, EdTech Insider listeners. It's very rare that we get such a special guest who's been not only a friend of the pod, but a friend of mine for so long.
Gavin Cooney, CEO of Learnocity. Welcome to EdTech Insiders.
[00:02:23] Gavin Cooney: Thanks for having me.
[00:02:24] Ben Kornell: Before we dive in, I think it would be great for everyone who may not know about Learnocity. learnocity. com A little bit about the company history and your reach, because it's so incredible how you've built this company and the way in which you're impacting students around the world.
[00:02:39] Gavin Cooney: Yeah. I mean, I always talk about Learnocity as the biggest ed tech company you've never heard of. We're powering a bunch of all the products that you know about and have bigger brands and we've got 40 million users. Last year we delivered 19 billion. That's billion with a B questions. Um, and. Yeah, we're, we're, we're the kind of underpinnings of a lot of the major products out there.
Um, we're really proud to have the impact that we do.
[00:03:06] Ben Kornell: Well, we're here to talk a little bit about some of your new AI features, and there's really no better place to start with AI than assessment, because it's incredible the way that data can unlock insights, but there's also ways in which We've never been able to implement assessments effectively because of time and labor intensity.
So first, let's talk a little bit about your AI tools and how you've oriented your essay scoring to save teachers time each week.
[00:03:38] Gavin Cooney: Yeah, thanks. So look, we all know that teachers are way overworked. There's a huge kind of teacher shortage now. There's a lot of people leaving the profession. Unfortunately, it's the most.
Noble profession, right? And people are leaving it because it's so hard. I'm in a Facebook group of Irish teachers, there's thousands of them, who are, it's called like Irish teachers leaving teaching. And there's thousands of them. And they're all complaining about the same things, about that overwork, and the amount of time, specifically, that they spend outside of the classroom.
So, you know, there's been surveys, we did some ourselves, and we saw some independent surveys, of teachers spending about five hours a week grading homework. And AI just comes in and it just can do so much better and so AI comes in at a really affordable level that can give much better feedback than ever would have been possible.
Much more verbose, actionable feedback for a student and can do that in a fraction of the time for a teacher. And what it really does is it enables types of assessments that were kind of too daunting and too time consuming for a teacher. For teachers before a teacher isn't going to assign a three minute speech to a class of 30 students when they're gonna have to listen to 90 minutes of say video and then give another 90 minutes of feedback.
It's just next to impossible to do on a normal night for a teacher, but they can do that if it's going to be auto graded and they can get that feedback. Much more feedback than they ever would have been able to write themselves. So I think it kind of opens up a kind of way more authentic assessment and what we wanted to do with this was we wanted to do a kind of a step change from what was available before.
AISA scoring has been around for a couple of decades and. What it was before the old fashioned AI, if that's not an oxymoron, old fashioned AI, what it meant was you would do, say you get a thousand responses for one essay, you would professionally grade them, and then you'd have a thousand responses and a thousand scores, and you'd build an AI model.
And 1001st essay could be auto scored reasonably effectively, but you've done a massive job there and it costs ultimately cost like a dollar an essay, right? And it's great for college board doing the AP exams or some like statewide examination when you've take hundreds of thousands possibly of the same essay being done, but it doesn't work.
It's too expensive for start for say formative and it doesn't work from essay number one. It works from essay number 1001. So we came along. Uh, with this, it works from essay number one, and it is one cent an essay. So it's kind of two orders of magnitude cheaper than what was available before, which kind of opens up a whole array of use cases that weren't there before.
And now it's kind of affordable to have a grade just homework and really enables teachers to assign just basic tasks. Better homework than they ever would have done before, better assessment than they ever would have done before with more authentic feedback, which is just amazing. I'm so excited to bring that to the market.
[00:06:48] Ben Kornell: So can you tell me a little bit about what makes your technology and your AI product different than just like a chat GPT wrapper or something like that? Like what's really going on under the hood? And then also like how, how do you see Lernocity differentiating your offering?
[00:07:07] Gavin Cooney: Like a lot of things now, it's built on LLMs, but it's built as a kind of an integrated solution to plug into it.
Our business is powering all the people's products, so it's an integrated solution that somebody, a product builder can go and build into their product. We're not selling this directly to teachers or to schools, we're selling to product builders who then in turn build products. And what we've done is we've taken a lot of data science to this and we have Gone and we've iterated and made this better and better.
So, there's a thing called a quick score, a quadratic weighted kappa. So, if you're hired to, you know, grade statewide examinations somewhere, what they'll do is they'll give you a hundred sample, Papers to grade, then they'll look at what you graded them as compared to what the official grade is, the kind of committee who has graded them kind of professionally, and they'll do a math calculation to kind of work out how you align to that, how close your grades to the official grade is, and that's called a quadratic weighted kappa, that alignment score.
A one is a perfect score, which is impossible. You and I. Hired in the same, grading the same essays is going to get maybe a 0. 7 of a quadratic weighting cap, a quick score. Anything around a 7 is a good score for two humans to grade the same essay. And that's how you evaluate the efficacy here of this stuff.
So we've come along and looked at this. Sometimes we're getting as high as a 0. 91 in terms of quick score, but a median about a 0. 8. Which is bonkers good. It's really, really good. And for one cent, that's unheard of. And what we found is, as we tune this thing and as we engineer around it, we're getting better and better.
So we were probably at a 0. 5 back in June, 0. 6 something back in August, and up now at 0. 7 something. And 0. 8 because we're getting much better at doing it and how we do it is we've got 50, 000 essays from kind of official sources, either publicly available data sets or from our friends and colleagues and clients in the industry.
So then what we do is when we make a little tweak to our engine. We then run a true 50, 000 essays and see if it got a little bit better, a little bit more efficient, or a little bit worse. And then we do that a hundred times between releases and see how much better we can make this thing.
[00:09:32] Ben Kornell: Yeah, I think there's a question too, generally speaking, around, What is the efficacy bar that we need to hit and what is the efficiency bar that we can unlock and what is the relationship between those?
And I think, you know, what I'm hearing you say too, is that in some ways this is enhancing the role of the educator. This is, you know, by far, the AI is almost like an assistant, not the be all and end all. And I find that so many folks are intimidated by AI as like taking jobs or undercutting processes.
But you're literally talking about places where students aren't getting feedback in a timely way. Educators are overwhelmed with the amount of grading and assessing. And this actually like optimizes the magic of the educator and their connection with, with kids.
[00:10:25] Gavin Cooney: That's it, right? So we call the product Feedback Aid.
Feedback, because it's about feedback, not necessarily about scoring, but that actionable, like a few sentences of description, trait by trait, of how the student can improve their writing going forward, because it's a formative tool. And an aid, because it's an aid. to a teacher. It's not replacing the teacher at all.
I'd be appalled if I just thought this was just giving out grades and nothing else. This is for feedback and it's to assist the teacher and make the teacher's life much easier. And yeah, there's loads of numbers around it and loads of how efficient it is and so on. But what you're really competing with is a massively overworked teacher who might be able to give like a thumbs up or a seven out of 10 or some basic bit of feedback.
A week after the student writes the essay here, there's much better feedback, much, much better feedback, much more detailed scoring and much quicker as well. So you're kind of competing against the norm and the norm is pretty bad to be honest with you. So, you know, I'm really excited about how we can raise that bar there.
This isn't supposed to be grading high stakes, super valuable, summative assessments. This is to massively improve. What's been done in informative.
[00:11:42] Ben Kornell: So, oh, that's resonates with me just to pivot a little bit. Are you talking to any of the LLMs and selling like some of the assessment data or, or insights you've learned?
Because I think the other thing we're seeing is. Large language model providers actually see AI in education as a key use case. But what they're really missing is high quality data sets that actually help train their, their platforms. Or do you think they're trying to go it alone? Like what's, what's your read on that?
[00:12:16] Gavin Cooney: So, well, so firstly, no, we don't train the black box at all explicitly and contractually. We don't do that. So we have, we're working with different LLMs. We have a multiple LLMs. Some of the models involve going to two LLMs, getting a score and feedback from both of them, working out which one is better.
Cause we know that like. Xllm is better at sentence structure and Yllm is better at content, for example, so we go and we do that. We work with the LLMs, obviously, but what we do is we work with the enterprise versions of those LLMs, so we're explicitly and again, as I said, contractually not training the LLM, not training the black box at all, and that's something we can kind of pass on.
That security, that data integrity is kind of part of what we're doing and part of the pilot we're doing. So we're not doing that at all. And yeah, we have a bunch of data. We have a bunch of data, but that's not our data to give away. And I think our customers would be hugely uncomfortable if they thought that we were doing something untoward with their data.
We're honest and sensible guardians of someone else's data. That's what we're here for. We're not doing anything silly on the side. It's like, it's a really kind of old fashioned business model when you think about it. For every dollar you give me, I'll give you 10 of value back in terms of the functionality I bring and what you would have to build instead.
And there's not like an ulterior motive where I'm doing something weird with data or selling AdWords to your students or something silly like that. It's just a kind of old fashioned. For every dollar you bring, we'll bring a bunch of value and it's kind of that simple and not trading the black boxes is part of that value.
[00:13:54] Ben Kornell: Yeah, this is going to be an interesting world that we move forward into where you have LLMs that have ingested huge amounts of information, but then you have specialized. Players like yourself who really understand pedagogy assessment understands psychometric data in a way that none of those LLMs can really ever replicate.
And you've got years of experience tuning that. And so, you know, will our buyers in the space understand enough the difference in quality? To buy the specialized and high quality products versus the non. And then even more interesting is there's a layer because Gavin, you power so many of these platforms.
So it's often not even an end school district. Or teacher that's saying, I want Lernocity in my product. They're working with products that leverage Lernocity. How do you think about the market forces and how we get the market to understand the quality discrepancies and what they might get from a generalized LLM versus specialized?
[00:15:02] Gavin Cooney: Well, look, I did this generalize tools in general. And then what we've been able to do is kind of go a mile deep on an inch wide on on certain functionality. So, any kind of LLM is just doing, I mean, their it's fascinating what a, what an era of technology we're in. And isn't it amazing, but they're like, they're.
Good at everything. And they're not particularly great at any one thing and to build something and to be able to go a mile deep and really focus on this and have our data science team work on that and improve that and tweak that and tune that, that really works and that's where we add our value, right?
I building on top of very, very good LLMs that are out there that just have a broader, a broader base. I'm sure there's an analogy about selling. A car versus selling a sports car, you know, there's, there's some things that are or a delivery van or whatever, like something that has a specific purpose as opposed to just a generic, a generic product.
[00:15:59] Ben Kornell: Yeah, it's making me think about Alex Arlen's article about the learning layer. In the future AI stack. And it's so interesting that you have multiple agents in your platform, creating the responses. And then that's feeding up to a front end application that may also be using agents in other ways. And so the quality of the stack is really, it's a complex engineering problem that the end user is almost unaware of, like in the best situation, It's just working.
So I think that's, it's a great pull the curtain back moment.
[00:16:38] Gavin Cooney: And because we are, we're powering, by the way, lots of, of, of customers and lots of publishers and lots of learning platforms, they don't have to go and reinvent this wheel themselves. You know, you think about Moore's law, everything's getting.
Cheaper and more easily accessible, but really one of the things that's not getting cheaper and more easily accessible is developer talent and especially AI development talent. So if you're in a learning platform or an edtech product or a publisher, you have a very finite number of engineers. So do you want to have them reinvent a wheel of doing essay scoring or whatever?
Or do you want to kind of leave that to the experts, especially when we're doing a sent an essay and that's kind of. It's just, it's wildly cheap, right? So do you want to do that? Or do you want to kind of focus your developers where they'll really add value in your product and to your students? And that's the whole, the whole promise of Learnocity and what we've been doing that that kind of leveraged model where we'll do something and it'll reach more millions of students.
And it means we can kind of, we can invest more in any one thing and any one of our customers or any one publisher or any one edtech product could ever do.
[00:17:47] Ben Kornell: What's great about that is that was your brand promise even before like generative AI, I mean, 2017, we were a customer of yours and. It's the, the pedagogic value that you brought in those days was so differentiated.
And now you combine that with AI enablement and capability, your team of 20 AI experts that you're able to bring and dive in.
[00:18:10] Gavin Cooney: Yeah, it's actually, it's actually more what we actually did was, and it's an interesting story. I know we kind of get into this. This is that. You know, ai, I wanna say AI came along, but there has been a revolution in ai, right?
And we got very, very excited about it. So when we started building, the first thing we built was AI authoring because there's 370 million questions authored in Learnosity. So there's a lot of people's job is to author questions in ity. So let's make that easier. So we built author aid against an aid to authors.
And to do that, we siloed off half a dozen or so developers and some product people. And we said, don't work on anything else. This isn't one thing in a complex, a roadmap with a bunch of dependencies. I want you in a room up by yourselves, working on this and not giving up and not sort of being pulled in different directions.
And it's interesting. What happened then was we went with a roadmap to ASU GSV this year. We had built some prototypes. We kind of whittled down a dream list of like 30 different things and said we're going to launch. These are five things we're going to launch that really add value to our clients, really save money for our clients, and are really kind of exciting.
Those things were author aid for authoring questions, feedback aid for grading essays, and Um, a D I checker to, um, to help with the authoring of questions and the maintenance of questions. A large scale item bank health check that's enabled to do large scale AI processing on an item bank. So you've got quarter of a million items.
And you want to check if they're still current, or if the images look like they're current, or if they need DEI checking, or to add to augment them or improve them in some way. It's kind of item bank health check. And the last thing then was a bunch of stuff around math, where bridging the paper digital divide, thereby allowing a handwriting recognition, generating math.
work solutions beside every question, doing author aid for math, doing multiple step math with feedback. So a bunch of math AI stuff, right? So those are the five things we went to ASU GSV with them. It says, Oh, like, let's talk to our clients and see what they think to say they're ripping my arm off.
Looking for it was, it was like that mean, like, give me like, take my money. That was the really wanted thing. these products. So on the way back, I'm flying back to, to Malta where I live and I've got a kid. I haven't slept in a week probably because I've been ASUGSV and I really should have slept. I stayed up all night on the plane and I wrote a memo to the board and to my exec team saying that we're going deep on this.
We're hiring 25, 25 more people in Dublin that are in AI labs on top of the half of those. Like eight people, I think, working on it at the time. And we're going all in on this. And we'll work out how to pay for it later. And sure enough, we, I landed, I sent the memo. There was not one person who thought that that was a bad idea.
And we're now, I think we've hired 15 out of the 25 people. We've dedicated another bunch of existing developers in that we opened up a new office to this in Dublin, and now we have a bunch of products, design, data science, and engineers all working on this AI stuff, and we're super, super excited, and the whole company is kind of invigorated by it.
[00:21:32] Ben Kornell: It's amazing, and I feel like we could talk for hours, because Just hearing about how you put that together so quickly. And then also the global infrastructure, the global reach, how do you convince a board to make a bet on something like that? When you're basically going off of the vision you see in the market, but everybody is like pulling back on investment.
It's like, there's so many layers here. I guess the last question I'd, I'd kind of end with for this conversation is Really, what does the future look like as you look forward for learnocity, for assessment, for AI enablement, what's making you like most optimistic and what is giving you cause for pessimism or, or pause?
[00:22:16] Gavin Cooney: Look, I think that this is the biggest kind of risk and threat to any ed tech company in a decade, right? It's a big change, the big thing, but it's the biggest opportunity for education in a hundred years. And I mean, I can sit here and I can talk about what I think is going to change, but like, we just know this change, right?
We know that there's a huge change agent out there. We find that people are looking, like, are looking to see what's coming out. People are excited about it. Publishers and their, and their customers, teachers know that this is here. And fundamentally, AI means that software should feel like magic and it should do stuff that just couldn't do before and I'd like to say the world is our lobster and, but not oyster obviously, the world is our lobster and you know, it's just such an exciting time to be in education.
All of this changes, and I don't know exactly how it's going to wind up, but like anywhere, there's change in anywhere. There's innovation. I'm here for it. I'm super excited about it. And I think that it's just such just kind of completely reinvigorated me and my team and our company and our products.
Just to kind of go after that and to sort of see how that changes and to be that innovation partner for people going forward. Cause we have that reach. Like if you and I then started a company tomorrow, we can certainly build a, a decent thing and we're bright guys and whatever it is, but we wouldn't have that kind of global reach.
Wouldn't have those 40 million users to students to go after that Learnocity has. We've just such a impossibly great. Position in the market to kind of go and to make hay while the sun is shining and this, and it's just such a great time to do it.
[00:24:02] Ben Kornell: And then you think about how that translates to what educators can do, what learners can do back to your, like, this is the hundred year wave that totally crashes down on the beach and can cause destruction, but it also reshapes the beach in a way that it never was before.
You know, there will be a few companies Catch that wave and ride the wave all the way in. And I, I, I'm betting on you guys at Ity. So was
[00:24:30] Gavin Cooney: that a, was that a hundred year storm, was that a, was that a reference to point break
[00:24:33] Ben Kornell: y you know, I
[00:24:35] Gavin Cooney: like your style. I like your style career Four,
[00:24:37] Ben Kornell: everyone's, yeah. Yeah.
But yes. Uh, in point break though, I get to beat Keanu Reeves. Just, I'm, I'm just saying. So . Okay. Um, so Gavin, if, if people want to find out more about what you're doing and what you're building at Learnosity, what's the best way for them to find out more?
[00:24:52] Gavin Cooney: Check out LearnOzzy. com, we have some of the webinars up there.
The one I did a couple of days ago to launch this feedback aid product, this round essay scoring is up there. Go have a look at it or reach out to me Gavin at LearnOzzy. com or follow me on LinkedIn and always up for a chat on these things.
[00:25:10] Ben Kornell: Wonderful. Thanks so much, Gavin from lunacity, a global powerhouse in AI, the biggest ed tech company you've never heard of.
Thanks so much for coming on ed tech insiders.
[00:25:24] Alex Sarlin: Thanks for listening to this episode of ed tech insiders. If you liked the podcast, remember to rate it and share it with others in the ed tech community. For those who want even more ed tech insider, subscribe to the free ed tech insiders newsletter on Substack.
