Closing the Attainment Gap: How AI Tutors Are Scaling Quality Math Education – Third Space Learning

Tom Hooper 00:00:22

So our goal, and it kind of links back to this point on mission, was making effective tutoring accessible to disadvantaged children at scale and so helping to close the attainment gap. The requirements for that did not change between human or AI tutor, but the scaling benefits do change with the latter. We need to show and make the case that this is a very effective means of providing spoken one to one tutoring that could not otherwise be afforded and distinguish that from, you know, working through games or digital content or, you know, equivalent. We also need to explain and teachers have a very good understanding of this, but nonetheless we always have to make ensure this is well understood, that our tutoring is embedded in school timetables. So it is a small part of the student school experience that supplements and reinforces their learning in class. So when people talk about humans in the loop, the teacher is the human in the loop, setting goals, booking sessions into the timetable and that loop, as it were, is one that sort of circles around the critical class and social learning experience at school.

Mark Taylor 00:01:40

Hello, my name is Mark Taylor and welcome to the Education on Far podcast. The place for creative and inspiring learning from around the world. Listen to teachers, parents and mentors share how they are supporting children to live their best authentic life and are proving to be a guiding light to us all. Today I'm delighted to be chatting to Tom Hooper from Third Space Learning. They are a one to one maths tutor for every pupil that needs it. Their voice based AI math tutoring is built by teachers for maximum impact on your pupils and staff. Now Third Space Learning has been involved with online tutoring in schools since 2013 and they've used all their knowledge to develop an AI tutor that is leading the way to support more pupils than ever. It's scalable for one, low fixed annual cost. For more details, check the show notes and use the links to see exactly what is possible. Hi Tom, thank you so much for joining us again on the Education on Far podcast. We were just chatting before that. We were sort of just about to be shut down properly. It was sort of November 2020 I think the last time we spoke and our worlds were going to look slightly different for a few. Thanks for coming back and I know our lives are very different now than they they were back then.

Tom Hooper 00:02:52

It's a pleasure Mark, thank you for having me back on. We seem to speak at moments of significant change, so it's a pleasure to be back on the show.

Mark Taylor 00:03:01

So why don't we do a little recap of of where we probably were then compared to where we are now. Because I think that will sort of flow nicely into, into the way sort of third space learning is, has changed somewhat. So back then 2020, what were you doing and how are you doing it? And then take us into how that sort of developed in the, in the meantime.

Tom Hooper 00:03:21

Yeah, brilliant. If I may, I'll actually step back slightly further than that to kind of explain how we led into that moment. So I founded the business Third Space Learning in 2013, so kind of an eternity ago in terms of the various policies and technology shifts that we've worked through, but the mission has stayed consistent. So to make effective high impact one to one tutoring accessible to disadvantaged children at scale, helping to close the attainment gap, we work with schools selling programs that are integrated in school timetables that teachers can direct to supplement and reinforce their class teaching coming into 2020. So when we spoke, just ahead of when we spoke, we were one of very few companies that actually did this. It was, you know, some tutoring was the thing that people did in home, kind of in the home market. We had worked, indeed have worked with thousands of schools, delivered millions of hours of treatment, have a huge amount of experience in this. When 2020 struck, we were a business that provided online human tutoring. We recruited STEM graduates in Asia to transform the supply of one to one human tutoring, increasing scale, lowering cost, increasing access for children in need. So we were very experienced and I think very capable and proven at delivering on that mission in a very innovative way. When the pandemic hit, we grew very quickly, helping many, many more schools and students in the UK to help catch up with the learning loss that occurred through, through schools being closed and indeed the attainment gap widening. Fast forward to today and the big change that we stepped through is shifting from a model where we used human tutors to a model where we now use artificial intelligence to provide live spoken, one to one tutoring that reflects all of the research and best practices in high impact tutoring, that leverages all of the experience and data that we've built over the last 13 years and completely transforms our capability to deliver on our mission of making one to one tutoring accessible to disadvantaged children at scale and so helping to close the attainment gap. And we now find ourselves in a world where we can significantly reduce the cost for schools. We can offer schools unlimited spoken one to one AI tutoring. Every child can have as much tutoring as they need, all directed by their class teacher. So it's, it's been a huge step forward in terms of delivery on that mission since 2020. But as you can imagine, a lot of twists and turns in that time as well.

Mark Taylor 00:06:18

So I guess the first thing that strikes me, and I'm sure many people listening, is you said about the spoken interaction from an AI point of view, take us into how that is integral and important in terms of what you're delivering and also what the AI does in relation to delivering your product, as opposed to it sort of generating its own information which it then provides for the, for the tutoring itself.

Tom Hooper 00:06:47

Yeah, very good questions. So there are at a high level, there are many AI tutors out there in the world right now that provide effectively homework help. They allow children to ask a question and get a good answer as quickly as possible. That is not effective tutoring. So effective tutoring is what we call a proactive agent. So an AI tutor that will lead a student through their learning toward the goal the teacher set over a period of weeks in particular, designed to work with children who don't have confidence in maths, who wouldn't know what question to ask, who would not put their hand up in class. So the design for effective AI tutoring is, or the criteria for that are very, very clear. And when we set out down this path to initially just to understand how does this work and can this work, we were very clear that our goal was not to sort of build an AI tutor that was a version of a large language model. Our goal was to build upon the evidence base and the research and the best practice for what effective tutoring was, but transform access to that with artificial intelligence. And that research base is very clear. It needs to be spoken and one to one, it needs to be directed by it. Tutoring needs to be directed by class teachers embedded in school timetables, high dosage, two to three sessions a week with student and tutor working through high quality content with formative assessment for learning and summative assessment for teachers. So the sorts of programs that we had always delivered, but understanding how we could deliver those with the scaling benefits of AI, and much of this in truth is common sense. If you came to this and said, what do I think quality tutoring would look like, you would say those things. But particularly post pandemic, there is a wealth of research in both the UK and the US that backs this up as well. So our goal, and it kind of links back to this point on mission, was making effective tutoring accessible to disadvantaged children at scale and so helping to close the attainment gap. The requirements for that did not change between human or AI tutor, but the scaling benefits do Change with the latter.

Mark Taylor 00:09:21

So how did that change really work for you? Because I guess you're managing, like you say, multiple tutors from all over the world and then you're suddenly thinking the AI model, Then you're talking about technology, you're talking about developers, you're working in a way that you may not have worked before. So talk me sort of through that sort of journey.

Tom Hooper 00:09:43

Yeah, it was at a number of levels. It was a very complicated journey, strategically, technically, operationally. And as a leader with the team as well, we first. It was in 2023 that we first started experimenting with large language models to see, at the time, to better understand them. There was a huge amount of hype around AI. Now there was, I think perhaps it was more confined, but within the sort of sectors we operate in, there was a lot of hype back then. And initially I was quite cynical about this. There's a lot of talk about large language models providing an AI tutor for every child. I think I probably made made the mistake a lot of people do, where my understanding of the technologies involved was reflected in my experience of ChatGPT. And my initial take was the sorts of children we work with and we're looking to help, they will not benefit from chat tutoring, where they ask a question and then get an answer, which I think was too cynical and defensive an approach. We. So in autumn of 2023, we thought, look, if we're going to be true to our mission of focusing on innovations transform access, we should be experimenting with this. If anyone should be, it should be us. And then secondly, as we started to look at this, our product hypothesis was given how we built our technology, which was in order to scale the human intelligence of thousands of graduates on the other side of the world, graduates, undergraduates on the other side of the world, in principle, that should give us a very good platform from which to scale artificial intelligence in a. With similar controls. And a lot of that. The logic behind that is grounded in the way we build our curriculum models. So as a teacher sets a goal and we use assessment to identify students learning gaps in order to achieve that goal and then create a program of learning, a sequence of learning objectives which a human tutor would guide a student through and which would scaffold that human intelligence, that human tutor's intelligence, to deliver that program in a consistent and measurable way that scaffold that architecture. This comes back to the hypothesis was before logically that should work for artificial intelligence. And this. This kind of keep coming back to this point, but it's been it's been the consistent through all of these decisions was we have built our systems like that in that way to deliver on the mission of transforming scale, lowering cost, increasing access to tutoring. And so there was a history through 10, 12 years of decision making to deliver that mission. And those decisions suddenly became, were amplified as these artificial, these technologies became prevalent. As we started to experiment with these in, in late 2023, at that time we're thinking let's just understand the potential for this. At this stage we, we quickly gained confidence that our sort of product hypothesis was true and very quickly started to get this into schools. So initially I think it was five schools, five London primary schools we were testing in from I think it's Easter 2024, just every week in those schools, working with the teachers, working with the students, getting feedback. And I think at that point in time we thought this is probably a three year thing and that will be quicker than the market. The market thinks it's a sort of five, you know, it's the future but that's at least five years away. We think it's the future but it's three years away. But as we started working through 2024, as I said week in, week out, increasing the number of sessions we're delivering and with the benefit of being able to benchmark the data from students working with our AI tutor within the same platform as the data from students working with our human tutors, we could quite quickly see that the two kind of main KPIs we were looking at assessment data so the post session quiz score which students complete after every session. By December of 2024 the students working with our AI tutor were achieving equivalent scores. I think it was 68% with AI tutor and 67% with our human tutors. And similarly we saw attendance rates were the same. I think it's 85% and 87%. And within this we could see just how quickly the technologies were developing and kind of correspondingly our ability to implement those technologies and move, ensure that our technology roadmap moved as quickly as the core generative artificial intelligence technologies that are now abundant. At the same time we could see that schools ability to afford high impact human tutoring in the UK and the US was rapidly diminishing. School budgets are in an incredibly challenging position in the UK and in the US where we operate. And so it felt obvious to us that given the mission, given the market challenge of the attainment gap and school funding, given the data we had as at December 24 and the pace of change of improvement. It seemed clear and obvious that we should focus all of our energy on developing best in class AI tutoring. That reflected the research. Best practices was co designed with schools. That's what we should do. And it sounds like quite a big gamble. It wasn't. This was based on 18 months of testing, an abundance of data and a very clear understanding of schools, what works, what they need, what their budgets are. So we took the decision to close down human tutoring and to as I said, go all in. Providing unlimited one to one spoken tutoring to schools. One set annual price £5,000 for UK primary school. Every child can have as much spoken one to one tutoring as they need. Cross year groups reflecting all the research and best practice that was the solution that would help our customers the most. That's what we were going to do.

Mark Taylor 00:17:01

And I think what you said there about it sounds like a gamble. I think if you'd asked me that sort of 20 minutes ago as we were sort of starting to speak, based on my limited knowledge of AI and how it works in my life and what my idea of people's perception is, it would sound like that. But because you've taken us through this wonderfully painted journey of that like you say, working with the schools, working with pupils, getting all that feedback, getting all those results back, it sounds like a very organic way and like you say maybe faster than you originally thought but it's it. I think the picture painted is a very clear one and like you say, if it means that as of that moment you, you decided to do it, you were going to suddenly help more pupils and help more schools actually be able to say well we know if, especially if it's a one off fee and we can give as much as we need to the people that really need it and we can know we can start to build our, the way that we utilize it in a different way than maybe we did before because we weren't quite sure where this was going to end. From a financial point of view then that really is a win win for everybody.

Tom Hooper 00:18:05

Yeah, we, I think we actually when we came to the decision we felt the biggest risk was not doing this rather than doing this. And we are, we're eyes wide open on this. We, this is the beginning of the road. We don't that there is a lot to improve but when we look at the data we are seeing improvements month on month on month. A month it did at the moment, week on week on week on week. So yeah there is a huge amount more to come in terms of what we're doing and the pace of changes is improving. So we feel very confident, again, looking at the data, looking at the feedback from teachers and students. We feel very confident that we made the right decision based on our mission, but based on what our customers needs are. Our customers budgets are how they want to help the children they want to help and how many of those children there are. This was the right decision.

Mark Taylor 00:19:06

And in terms of that feedback that you got, in terms of interacting with an AI tutor rather than a personal tutor, what were some of the things that was said above and beyond? It's. It's the same or it's fine or it works for me. Is there any sort of particular changes which weren't good or bad, but just an obvious thing that people sort of mentioned that you sort of took on board as it was being developed?

Tom Hooper 00:19:29

Yeah, I think there were two main themes of discussion, particularly through 2024, and these are. These have continued into 2025. But we were. I mean, I was speaking to teachers every day about this, but in the broadest sense, their feelings and attitudes about AI in general and education. But then I'll see more specifically about tutoring. And I think that the thing that really came across in 2024, as we were building up toward this decision was teachers were not. There was a sort of a perception in the sector, a feeling in the sector. The teachers were nervous about AI. That was not what we were. We were seeing at all. And this was from dozens and dozens and dozens of conversations. Teachers are saying, our job is incredibly hard, we're under a huge amount of pressure, budgets are a disaster. So we believe that is a huge opportunity in terms of helping us to solve these fundamental challenges. So we see that, but we want to make sure that we're working with people who understand our reality and. And have a deep understanding of education and a trust to develop that. And I think we've worked with thousands of schools in 2013. I think we started within tutoring. We absolutely bring that to the table. So there was a feeling of opportunity against the context of challenge from pretty much every teacher we spoke to. So we felt confident that there was a willingness for this to work in the teachers we spoke to at a student level, the main. I suppose the question we had coming into this was, will a child be able to engage with an AI tutor? Will there be something fundamentally different of sort of technical. I don't know what you call, like an emotional block that they wouldn't build before? I think that that was. It remains a very important question. But, but that was a particularly important question at the beginning of this journey. And we, we had a hypothesis, I certainly had a hypothesis back then that, that actually we would see children would, would, would achieve and have a sense of agency and ownership with an AI tutor that they would not have with a human tutor. And let me explain that for you. One of the things that is particularly effective with online spoken one to one tutoring is giving children the chance to ask questions and vocalize thoughts, misconceptions, explore ideas that they might be too. They might lack the confidence to do in class, they might lack the confidence to put their hand up to ask in front of their peers, often down to confidence. And what we saw in online tutoring with human tutoring time and time again with teachers saying that child puts on the headset and starts talking away to the math tutor, this is magic, they don't do that in class for all the obvious reasons we know about. And I suppose our hypothesis was actually that that might be amplified with an AI tutor because it is a low stakes environment where children don't feel judged but do have the ability to talk through the point the concepts they struggle with in class on a regular basis in a structured way. And that because that relationship with an tutor can be longer term and more owned by the student than with a human tutor, which is necessarily short term because of the cost that that might give the student a greater feeling of agency over their attitude in a positive way for their learning. And we have, I'm going to talk anecdotally here. We have definitely seen that it's more than anecdotally. We commissioned research, we commissioned Educate Ventures to do a review of all of our data and to interview teachers and students. That was one of the key things they pulled out as well of children feeling they had sort of benefiting from that low stakes environment where they could ask the sorts of questions they might struggle with, they wouldn't feel charged and they could talk through these concepts. So I think the, I think that was one of the themes of understanding that also gave us a lot of confidence that actually the dynamic between a child and an AI tutor could be a very positive one, reflecting the sorts of learning experiences that spoken one to one tutoring can provide.

Mark Taylor 00:24:50

And from the outside this might be because I've had a conversation already today about games and coding and gamification and that kind of thing. But the reason it brings, brings it back to mind is the fact that children spend a lot of time doing computer games, being online, actually enjoying the process of learning in their own way with a non human entity, so to speak. Whether that's just through gaming, whether it's through. However the, the, the development of any journey within any game happens to be. So I can understand that sort of feeling, feeling supported and, and in a familiar environment with I'm, I'm online, I'm learning, I'm doing it in a way well, I say where that sort of low barrier happens to be. But it's probably maybe more familiar than maybe you would think as this is an AI tutor, maybe it doesn't feel quite so much alien to them, if that kind of makes sense.

Tom Hooper 00:25:43

Yes. And so there's a lot of things going on there in what you just said. So the, the learning experience has to be engaging and you want to reward students for hard work progression. At the same time, learning is hard and you can't shy away from that. So it can't be too much about reward and engagement. It also has to get into the difficult things which students might find hard. So there's an appropriate balance between the two. There can also be and quite understandably a nervousness from teachers and parents about students spending too much time online. And I think we need to balance that both. We need to show and make the case that this is a very effective means of providing spoken one to one tutoring that could not otherwise be afforded and distinguish that from, you know, working through games or digital content or equivalent. We also need to explain, and teachers have a very good understanding of this, but nonetheless we always have to make sure this is well understood, that our tutoring is embedded in school timetables. So it is a small part of the student's school experience that supplements and reinforces their learning in class. So when people talk about humans in loop, that the teacher is the human in the loop, setting goals, booking sessions into the timetable and that that loop, as it were, is one that sort of circles around the critical class and social learning experience at school. So this is, it's not the case of a kind of a parallel distinct learning stream where students sit with their tutor. It is success is in harmony with school, with class, with social learning, and ensuring children have a one to one experience and that every child can have a one to one experience that can build their confidence so they can engage in the fullest possible way in class, with their teacher, with their peers. And that is a very healthy picture that we play a small but I think very valuable part in.

Mark Taylor 00:28:15

And in terms of, of this model, you've really explained it really clearly about how it works from this sort of maths point of view, which of course is your speciality. So are you seeing it replicated in the same kind of way across different subject areas or. Or sort of the same sort of format in different areas? And are you sort of able to support each other and work in that way, or is it very much kind of depending on the ethos of the company organization to begin with, develop what they need.

Tom Hooper 00:28:44

So what. What we haven't seen in either the UK or the US is other companies delivering high impact tutoring with an AI tutor. And it's very important to be specific about what sort of tutoring we're focused on and why, and the kind of counterfactual to this. I'll explain briefly. There is a ton of kind of AI tutors, learning assistants out there, you know, versions of large language models where you ask it a question, it gives you a good answer quickly, it can help you with your homework, it can help you with practice. That can be good. That is not what we're doing. That is not high impact tutoring. There is a wealth of evidence that shows that delivers little learning gains, in particular for those children with the greatest need. So that is not our focus. Our focus is what does the research and evidence show that has the highest impact on students, in particular, students who need the most help, which is high impact tutoring, which is a much more complex beast, because the sorts of agents you have to build are proactive agents that would lead a student through their learning over time. An agent that is not waiting for the student to ask a question that it can give an answer to. And that is a. There is a lot more complexity in that. Similarly, there's a lot more complexity in delivering effective spoken one to one AI tutoring inside noisy classrooms with children that murmur and think out loud and talk to their mate and talk to their teacher and there's lots of background noise. That is a very complex environment. And understanding how to do that operationally and technically and from a user experience point of view and from hardware point of view, that is very, very difficult. Whilst we have two years of experience with air tutoring, now, we have 13 years of experience of implementing that in classrooms with human tutors. So I think the operational and technical complexity of spoken high impact tutoring is significant, which is why, as yet, we've not seen many other providers starting to do it. There are a couple who do it in the US from a reading point of view, but we've not come across anyone else doing it from a math point of view in terms of what we intend to do when we look to next year 2026, we intend to focus on two new areas. Number one will be testing out English programs in UK primary schools, so particularly focused on reading phonics and comprehension. So we hope to get that in schools later this academic year. And then in the US we're going to focus on math in Spanish for Hispanic students. So 29% of students in US schools are Hispanic. High failure rates, high achievement gap for Hispanic students. And so providing versions of our math programs that can cater to their linguistic needs and learning needs is something that we also intend to get into schools and start learning from in calendar year the first half of 2026. So that the the sort of general abilities to expand into new subjects and languages with an tutor are you can move much, much quicker. And the tooling and infrastructure that we've built to launch our maths in English programs will allow us it's subject and language agnostic. So it will allow us to move much more quickly into into those areas.

Mark Taylor 00:32:34

And this might be too simplistic, but it does seem to me that understanding how like say individual companies, how individual organizations and what they claim to do is really important to look at the exact detail like you just explained in terms of what you deliver. And I sort of half think of this in terms of you can talk about the Internet and people go okay, there's lots of things going on there. And then also I have these particular apps that I use that I know are trusted and they're built by these people and they work in this way. And I'm happy to use that because that's part of my life. And I like to say the difference between the language models and having a speed specific company like yourself be able to deliver it, I think sort of has that sort of harmony based thing. Certainly in my use of AI it's that things which are built into systems and workflows and apps and things that I use I know are very, very important because they work specifically for the things that I need as opposed to just going on to chat GPT or Claude or just asking a random thing. So I think, I think that sort of personal relationship and, and that understanding I think is in that basic model is sort of one when I sort of see sort of through those eyes a little bit.

Tom Hooper 00:33:47

When we first met mark in 2020, a big thing that's changed in schools and the same in the UK and the US since then is there is a much keener understanding of what effective evidence based tutoring looks like. So both the UK government and the US Government put significant sums of money behind implementing or funding and researching and explaining what effective tutoring looks like and thereby steering the market toward that. So the NTP was much, much more specific than kind of what they were doing with asset funding in the US Although in the US there was much, much more funding behind it. The NTP was very, very prescriptive. It was founded in part by the Education Endowment foundation in saying this is what good looks like. Therefore this providers of these sorts of programs are the only people who will be able to participate in the national tutoring program. And therefore school leaders had a very clear understanding of what good looks like, you know, and it can. These are educators. It reflects their understanding and their education and their training and common sense. So I think when we now come to market with our history and explaining our AI tutoring within that language and that framework and that evidence base, we're kind of moving with the grain of how school leaders understand effective strategic interventions for their children.

Mark Taylor 00:35:29

And going forward, you sort of mentioned in terms of obviously English and the Hispanic kind of focus you can do in the US From a technical point of view, do you just see the expansion as a gradual one that you can ease out and affect more and more people, or is it more the fact that the technology is going to become so advanced and the agents are going to become better and better sort of month on month, year on year, that it can just become even more personalized and just enable each pupil to get more out of what you're already able to deliver?

Tom Hooper 00:36:07

Good question. I think it's both. Maybe I'll answer the question in this way. I think, as it seems from our vantage point, very clear that within the next five years, I think every school will provide an air tutor to every child within that school that will support that child throughout their school career, helping them work through the content that they're doing in class, the curriculum that they're covering through that term and that year, and embedded in their school's reporting systems. So, you know, and reflecting the sorts of research and best practices we're talking to, I think that will become a core piece of software that every school will provide or purchase and provide like a management information system. It's kind of the counterfactual of them not doing that. Given what we know today and assuming the trajectory of the sort of rate of improvement we know exists, the counterfactual seems odder to imagine than. And that vision for how schools will work. And that is incredible. It really is. And we're providing a version of that Future today. For £5,000 every child can have as much spoken one to one high impact tutoring as is needed embedded in the school timetable, reinforcing class teaching, accessible at home, structured evidence reporting back to teachers. So that I think that's how we see the future when we know that I mentioned English within primary schools. I think 89% of tutoring within the NTP was maths or English. Those are the two kind of key high stakes exams, fundamental subjects that students have got a master for. Both kind of long term educational but also professional and social development. So it seems obvious to us at primary that that's what we should focus on. And if we could say to schools, every child can have a high quality, unlimited one to one tutor for maths and English to support their classwork, to help them at home and to ensure they succeed and to avoid those gaps building up from year four, year three, year five through year six, that would be an amazing thing. Our goal is to try and have that in place in time for 26, 27. So it's both kind of what, where does the problem exist and kind of what does the technology enable us to do and how quickly. But I think that's how we see the world and how we see what we offer and our place within that development.

Mark Taylor 00:39:01

Yeah. And I can, I can see that working really, really well because I think even for someone like myself. So if I think all the way back to when I was in school there was certain key parts of, let's take it as mass is the example where even if I wasn't necessarily having extra tutoring, it would have been very helpful maybe to be able to have, have a virtual tutor or to have an AI tutor that I could then go into to cover certain things which were related to things I could catch up with based on what my teacher could say or you know, let's just do a little bit of this, you've got the ability to go into here, find the information, do some practice on it, support what you need to do and then bring it back. So I like the fact that maybe is the future developed so you've just got more access in an easier way to be more supportive and maybe take a bit more ownership. Maybe I guess that's possibility, whereas I wouldn't have been able to have done.

Tom Hooper 00:39:55

That before and whilst controlled by teachers. So one of the broader debates within education is that the challenge of every child can go home and can just ask ChatGPT to do their Homework, that's really, really difficult. So how can you, how can schools, how can children use AI as they. They will, they will. You're not going to stop that. But in a constructive and controlled learning format. And I think that the sorts of things we're building out with teachers will allow that. So if a child is struggling with their homework or potentially their homework is working through a lesson plan with the AI tutor, so they have the support and access of AI, but within the control and structure of a lesson plan that the teacher set and the reports back to the teacher. So I think there are, there are kind of what we're building and indeed how it's being used today can help with some of those challenges whilst retaining the reality of the benefits for children where they will be using it.

Mark Taylor 00:40:58

Yeah, I think it's fantastic and it's a really exciting prospect. And I think, like you say, there are so many benefits from a workload point of view, a student point of view, a parental point of view, an overall school support point of view. And like I say, this is just the, the beginning of the journey, which is incredible to think, really. So we always seem, ask and get some ideas of what you think about our acronym of fire, Feedback, Inspiration, Resilience and Empowerment. What is it about one of those words, all of those words that sort of strikes you off the bat and maybe certainly within how you sort of see that through third space learning.

Tom Hooper 00:41:42

Okay, let's do this. I'm going to go with one for each word. I think we've been very fortunate and very grateful for the feedback we've had week in, week out from teachers and students over the last 18 months. As we've been in schools every week. I would not be here talking to you about the things we're doing today and the transformative impact that can have and will have without that feedback. So we've been very grateful for that kind of, you know, and equally, you asked at the beginning about the change we've worked through. It's, it's one thing to explain it in 30 seconds from a strategic point of view, but quite another thing to work through. And the resilience of our team has been unbelievable. And indeed, I'll go with inspiring. And a lot of those two together, it's been brilliant. And I think the. We also talked about the challenge teachers face. Their workload, financial pressures. It's a very, very difficult job, particularly right now. They articulated to us that they see, they can see how AI could play a role in squaring that circle for them. I like to think we're playing a small part in that opportunity, and I think that that reflects the empowerment that AI can give to teachers. And I hope we will be a great example of that in the weeks, months, years to come.

Mark Taylor 00:43:10

I love that. And I think my biggest takeaway today is the harmonious nature that you've been able to explain in terms of. So often we sort of have this concept of AI over here and sort of human interaction over here, and the picture you've been able to paint about using it in a productive way to support your mission and therefore to support pupils and people within education. But the conversations that you're having, the feedback you're having, the ongoing dialogue so that all of the new technologies and the human side of things are working hand to hand with an overarching understanding of those people in education, to be able to sort of do that for the best of everyone's ability in there, which I think is a. It's a perfect way to sort of round up what we've been talking about. And I'm just so glad to be able to share this and to get a sort of a real picture of what that looks like, because I'm sure so many people are unsure about when they hear AI in education, even if they're in education, let alone from looking from the outside in what that might look like. So, yeah, Tom, thank you so much. And I really do appreciate you sharing that journey with us.

Tom Hooper 00:44:12

Brilliant. Well, thank you for having me on, Mark. And thank you, everyone, for listening.

Mark Taylor 00:44:17

Education is not the filling of a pail, but the lighting of a fire.