Foundational Impact | Official Trailer

In this episode of Foundational Impact, host Daniel Emmerson speaks with Hannah-Beth Clark, Director of Product and former AI Quality and Safety Principal at Oak National Academy. Hannah-Beth shares insights into building AI tools designed specifically for teachers, including Aila, Oak's AI lesson planning assistant. She explores the mechanics of Retrieval-Augmented Generation (RAG) using curriculum-aligned content, the necessity of intentional friction in tool design, and how robust safety and auto-evaluation systems maintain high pedagogical standards while keeping human expertise at the centre. Key Topics & Timestamps00:00 Open Curriculum Access00:24 Show Introduction01:25 Oak Origins and Mission04:21 How Teachers Use Oak07:00 Designing Productive Friction08:48 How Aila Plans Lessons11:30 RAG and Safety Guardrails14:26 Hannah Beth Role17:53 Evaluating AI Quality22:19 When Experiments Fail24:05 Trust and Human Oversight26:00 Content Store and Open API28:40 Advice for AI Skeptics31:55 Closing ThanksKey Takeaways & HighlightsProductive Friction in Design: Tools should intentionally require input and critical thought from teachers rather than offering a one-click output, ensuring lessons remain tailored to individual classroom contexts. Retrieval-Augmented Generation (RAG): Grounding AI generation in a curated, high-quality corpus of national curriculum-aligned content ensures outputs meet strict pedagogical standards compared to general models. Nuanced Safety Guardrails: Automated content moderation must distinguish between genuinely toxic content and sensitive, age-appropriate educational topics (such as classic literature) that require specific guidance rather than outright blocking. Human Ground Truth in Evaluation: Automated evaluation tools can effectively flag technical issues like regionalisms or poor multiple-choice distractors, but expert human review remains essential to maintain output quality. Purpose-Led Integration: Schools should approach technology by identifying clear classroom challenges first, avoiding adoption merely for the sake of using AI. Resources MentionedOak National Academy: Free curriculum resources and lesson materials aligned with the English national curriculum. https://www.thenational.academy/ Aila: AI-powered lesson planning assistant developed by Oak National Academy. https://labs.thenational.academy/DfE Content Store: Department for Education repository providing structured educational data and open curriculum APIs for developers. https://aicontentstore.education.gov.uk/Guest: Hannah-Beth Clark (Director of Product, Oak National Academy) Hannah-Beth Clark is the Director of Product at Oak National Academy, having previously served as its AI Quality and Safety Principal. Prior to joining Oak, she was part of the founding team at Oasis Academy South Bank, where she served as a senior leader for ten years before becoming Headteacher in Enfield. She also serves as a governor for two primary schools in Cambridge.

In this episode of Foundational Impact recorded live at the Festival of Education 2026, host Daniel Emmerson speaks with Dina Foster, Head of Educator AI Training at Educate Ventures Research. Dina discusses how schools can transition from experimenting with AI to developing a coherent strategy. She emphasises the importance of AI literacy as a firm foundation for schools, the value of addressing existing priorities rather than treating AI as an extra burden, and the necessity of engaging all stakeholders in policy development. Dina also shares insights from her work guiding the Shape of the Future Coalition and supporting state schools through targeted AI leadership programmes.Key Topics & Timestamps00:00 Incorporating AI Into Existing Priorities00:42 Welcome To Foundational Impact And Festival Of Education01:38 Assessment In The Classroom Of 203602:40 Crowded Horizons For School Leaders05:14 Building Foundations With AI Literacy06:16 The Shape Of The Future Coalition08:35 AI Leadership Support Programme For State Schools09:27 Compliance And Safeguarding Concerns12:12 Collective Intelligence And Values14:21 Managing AI With Parents And Stakeholders17:21 Key Advice For School LeadersTakeaways & HighlightsAI As A Foundation: Building AI literacy is essential to weather future technological shifts, acting as a keystone rather than building on sand.Integrate, Do Not Add: AI should not be an isolated, extra silo; it must be incorporated into existing school priorities and used to solve current challenges.Community Support: The Shape of the Future Coalition offers a free, research-backed network for school leaders to share practices and stay updated on AI regulations and frameworks.Stakeholder Inclusion: Creating effective AI policy requires a holistic approach that includes teachers, leaders, governors, parents, and learners rather than focusing on just one group sequentially.Evidence-Based Decisions: Leveraging research expertise, such as the three decades of experience from Professor Rose Luckin, helps lend rigour to AI use cases and action research in schools.Resources MentionedEducate Ventures Research (EVR): Research consultancy led by Professor Rose Luckin. https://www.educateventures.com/Shape the Future Coalition: Free network for school leaders to share AI practice and research. https://www.shapethefuturecoalition.org/Human Intelligence First - The Tool is the Smallest Part: nine practitioner-led action research projects conducted over the 2025/26 school year by school and trust leaders from across England. https://www.educateventures.com/research-human-intelligence-first-the-tool-is-the-smallest-partAI Quality Mark: A thorough and rigorous framework for schools to work through their AI implementation. https://www.goodfuture.foundation/ai-markCollective Intelligence: A collaboration between Educate Ventures Research, Good Future Foundation, Cambridge OCR, and Big Education. https://www.collectiveintelligence.org.uk/Guest: Dina Foster (Head of Educator AI Training, Educate Ventures Research)Dina is an experienced former secondary school Deputy Head teacher and is currently was doctoral candidate at UCL with a research focus on how schools develop AI strategy to harness the benefits of AI, and augment teaching and learning. She gained her MA in Education and Technology and also holds a Master’s in Teaching, both from the Institute of Education. Dina is particularly interested in supporting school leaders with the development and implementation of successful EdTech strategies to address school priorities. In her current role at Educate Ventures Research, she heads the Shape of the Future Coalition, providing CPD training around AI and strategy work with school leaders.

In this episode from the Festival of Education 2026, host Daniel Emmerson speaks with Professor Miles Berry, Professor of Computing Education at the University of Roehampton. Miles discusses what AI changes for teaching and learning, the importance of universal AI literacy in national curriculum reforms, and how initial teacher training is evolving to address generative AI. He shares insights on balancing teacher efficiency with preparation quality, the ethical nuances of machine-generated feedback and academic integrity, and how tools like "vibe coding" are changing software development for educators.Key Topics & Timestamps00:00 Vibe Coding and Software Creation00:28 Host Introduction01:00 Live at the Festival of Education01:19 Modeling Ethical AI and Transparency02:54 AI Literacy and the Head, Hands, Heart Approach04:07 Curriculum Reform and Universal AI Literacy05:23 Teacher Training and Lesson Planning07:41 Quality vs. Efficiency in Planning10:16 AI Marking and Student Motivation12:28 Teacher Education Frameworks and SEND13:49 Multimodal Tools and Language Acquisition15:22 Roehampton’s AI Strategy for Graduates17:56 Handling AI Integrity in School Interviews20:02 Defining Cheating and Transparency Acknowledgements22:09 The Joy of Vibe CodingKey Takeaways & HighlightsHead, Hands, and Heart: Teaching AI requires a balanced framework—understanding how models work (head), practical prompt engineering and bias mitigation (hands), and analyzing societal and ethical implications (heart).Universal AI Literacy: AI education cannot be reserved solely for computer science students; upcoming curriculum reforms aim to ensure every young person develops core AI literacy.Preparation Over Output: While AI can generate lesson plans quickly, the true value of planning lies in teacher preparation and thinking through the lesson, which automated tools might inadvertently bypass.The Human Element in Feedback: Students often value feedback because it signals that a human teacher cared enough to review their effort; automated AI marking risks demotivating learners, especially in creative subjects.Embracing Vibe Coding: Natural language programming tools allow non-traditional developers to build tailored, useful educational software through simple, iterative conversations.Resources MentionedUniversity of Roehampton Secondary PGCE: Initial teacher training programmes for secondary education. https://www.roehampton.ac.uk/study/postgraduate-taught-courses/pgce-secondary/AI4K12 Project: Initiative providing resources and guidelines for AI education in K-12 schooling. https://ai4k12.org/BCS, The Chartered Institute for IT: England computing curriculum committee promoting digital skills and computing education. https://www.computingatschool.org.uk/digital-themes/cas-ai-zone/Guest Contact InformationGuest: Professor Miles Berry (Professor of Computing Education, University of Roehampton)Miles is Professor of Computing Education at the University of Roehampton, where he is head of curriculum and recruitment for the secondary PGCE as well as leading the PGCE in computing course. He supervises doctoral students and is a member of the University's AI development and innovation group. Prior to joining Roehampton, he spent 18 years in four schools, much of the time as an ICT coordinator and most recently as a head teacher. He has worked with Cambridge University Press and Assessment on a number of curriculum and training development projects.He helped write England's national curriculum computing programmes of study. He gave evidence to the House of Lords AI select committee, was an expert advisor to the US-based AI4K12 project, and supported Georgian universities in developing academic integrity policies. He chairs the BCS England computing curriculum committee, and until recently was chair of the National Centre for Computing Education's academic board.

In this episode live from the Festival of Education at Wellington College 2026, host Daniel Emmerson speaks with Grainne Hallahan, Head of Community at Teacher Tapp. Grainne shares unique data-driven insights into how UK teachers feel about artificial intelligence, the shifting patterns in subject-specific AI adoption, and why asking precise, time-bound questions is essential for capturing classroom realities. She also offers practical guidance for school leaders on policy formulation, teacher anonymity, and effective survey design.Key Topics & Timestamps00:00 Friday Behaviour Questions00:20 Introducing Grainne Hallahan01:49 What Is Teacher Tapp03:10 Designing Daily Questions and Finding Insights06:38 Panel Randomisation and Response Experiments10:22 Longitudinal Data and Behaviour Tracking11:13 Tracking Teacher AI Adoption Trends Over Time12:51 The 10 Percent Non-User Group and Environmental Concerns13:50 Subject Trends and Primary vs Secondary Usage16:42 Sharing Findings with Policymakers and the Media18:36 Teacher Motivation and Anonymity23:21 School AI Policies and Future Research Areas24:54 Tips for School Leaders Writing Effective SurveysTakeaways & HighlightsData-Driven Teacher Voice: Daily three-question surveys sent to a re-weighted panel of 10,000 teachers allow Teacher Tapp to capture real-time classroom sentiment and represent the profession accurately across government consultations and media broadcasts.Shifting AI Adoption Patterns: The proportion of teachers who have never used AI for schoolwork has fallen to under 10%. Primary teachers initially led adoption for generating exemplars and short-form questions, while maths teachers have seen a major turnaround since tools gained image and diagram generation capabilities.Environmental and Ethical Concerns: Environmental impact has emerged as a primary reason why the remaining minority of teachers choose to abstain from using AI tools in their professional duties.Anonymity Drives Honest Reporting: True data separation and absolute respondent anonymity are vital for understanding shadow AI usage and unearthing candid feedback that school policies might otherwise mask.Designing Actionable School Surveys: When surveying staff, school leaders should restrict questions to recent, specific timeframes (such as "this week"), keep options below ten, and eliminate unnecessary verbiage to obtain clear, reliable responses.Resources MentionedTeacher Tapp: Daily survey app for educators in England. https://teachertapp.com/Guest: Grainne Hallahan (Head of Community, Teacher Tapp) Grainne Hallahan is a teacher-turned-journalist who now sets the questions at Teacher Tapp. She is interested in all things education, but gets particularly excited about what keeps teachers teaching, and how changes to education policy can make a difference to those doing the hard work in the classroom and school leadership.

In this episode from the Festival of Education at Wellington College 2026, host Daniel Emmerson speaks with Dr Iro Konstantinou, Head of Research and Impact at Eton College. Dr Konstantinou discusses the role of artificial intelligence in education, highlighting findings from primary data gathered across diverse UK schools. She explores the concept of cognitive offloading, how students' perceptions of AI are influenced by their background and social media, and how schools can prepare students for an AI-impacted workplace by fostering critical thinking and transferable skills.Key Topics & Timestamps00:00 Cognitive Struggle and Early AI Perceptions00:28 Introducing Dr Iro Konstantinou01:23 Highlights from the Festival of Education02:00 Overview of Research at Eton College03:25 Gathering Primary Evidence on Student AI Use05:24 Regional Disparities and Student Attitudes Towards AI06:57 Personal AI Usage and Data Privacy Awareness08:24 The Role of Social Media in Shaping AI Perceptions09:19 Teachers' Concerns and the Cognitive Offloading Trap11:04 Classroom Strategies: Prohibition vs Socratic Prompting12:11 Workplace Anxiety and Changing University Choices14:27 Supporting Schools with Industry Skills and Best Practice15:25 Community Outreach and Collaboration at CIRL17:11 How to Connect and Get Involved with Eton’s ResearchKey Takeaways & HighlightsAcademic Skepticism vs. Overconfidence: Students at highly selective academic schools often express a desire to maintain "cognitive struggle" and avoid AI, whereas students in other settings use AI for personal tasks without fully grasping data privacy risks.The Influence of Social Media: Many young people rely on social media platforms rather than formal guidance to form their understanding of AI ethics, safety, and application.Diverse Classroom Approaches: Teacher responses range from strict prohibition to incorporating Socratic prompting, highlighting the need for explicit metacognitive training to support critical thinking.Shift in Workplace Expectations: Students are altering higher education paths due to career automation concerns, pointing to a need for schools to emphasize transferable human skills and philosophy-driven thinking.Democratising Educational Research: Outreach initiatives focus on sharing best practices and supporting schools in underserved areas to ensure AI evidence-based guidance is accessible across the sector.Resources MentionedThe Tony Little Centre for Innovation and Research in Learning (CIRL) at Eton College: Research hub focusing on pedagogy, evidence-informed practice, and AI impact. https://cirl.etoncollege.com/Department for Education (DfE) Gen AI Product Safety Standards: Standards addressing friction in learning and cognitive offloading risks. https://www.gov.uk/government/publications/generative-ai-product-safety-standardsGuest Contact InformationGuest: Dr Iro Konstantinou (Head of Research and Impact, Eton College)Dr Iro Konstantinou is an education researcher and adviser specialising in evidence-informed practice across schools, universities, and the wider education sector. She currently serves as Head of Research and Impact at Eton College, where she leads research on curriculum development, professional learning, equity, and the impact of AI on education. Previously, she held academic and leadership roles at Pearson, UCL, and the University of Warwick. She is also a trained coach, having completed an undergraduate course in coaching at the University of Cambridge.LinkedIn: https://www.linkedin.com/in/dr-iro-konstantinou/

Episode Overview In this episode from the Festival of Education at Wellington College 2026, host Daniel Emmerson speaks with Victoria Hedlund, founder of GenEd Labs. Victoria discusses bias in generative artificial intelligence, the importance of inclusion, and what meaningful human oversight should look like in education. She also explores how educators can challenge AI outputs, balance productivity with equity, and preserve the essential human element in teaching. Key Topics & Timestamps00:00 Putting Output on Trial00:33 Introduction to Victoria Hedlund01:37 Researching Bias and Human Oversight03:11 Why Schools Must Focus on Bias05:56 Sparks, Lightbulbs, and Gender Stereotypes08:12 Frequency of Exposure: Teachers vs AI09:09 Visual Representation and Image Bias13:06 Questioning Assumptions and Model Iteration17:09 Equity Versus Productivity20:49 The Human in the Loop24:30 Embracing Messy Humanness in EducationKey Takeaways & HighlightsPutting AI on Trial: Teachers and students should treat AI outputs critically, challenging assumptions and testing validity rather than accepting generated content at face value. Deficit vs Strengths-Based Framing: Standard AI prompts often apply deficit-based language to neurodivergent or marginalised learners, requiring specific contextual framing from educators to deliver equitable outputs. Frequency of Exposure: While human teachers hold natural biases, students interact with AI models through hundreds of continuous iterations, amplifying potential harm if outputs remain unmitigated. Average vs Instantaneous Productivity: Relying on AI for quick administrative wins can sacrifice long-term accuracy, making upfront contextual alignment crucial for meaningful workload reduction. Preserving the Human Element: Complex emotional needs, non-verbal communication, and relational learning cannot be replicated by software, keeping human teachers central to classroom learning. Resources MentionedGenEd Labs: Innovation lab focused on artificial intelligence research, inclusion, and educational tools. https://www.genedlabs.com/Guest Contact InformationGuest: Victoria Hedlund (Founder, GenEd Labs)Victoria is a thought leader, researcher and consultant focused on fair AI use in education. Known as the ‘AI Bias Girl,’ she creates practical awareness and mitigation strategies for educators and leaders, using her research of emergent AI bias in educational contexts. She is a major voice in AI pedagogical alignment and has contributed to the updated DfE and Chartered College Safe and Effective Use of Generative AI in Education Guidance, and to the British Computing Society's AI course ‘AI Confidence’. She is editor of AI Bias in Education and author of 100 Quick GenAI Prompts for Teachers and Educators. She sits on the advisory board for the Institute of AI Education, is a member of the Centre for Digital Information Literacy in Schools Expert Working Group, and is part of Oxford University's AIEOU network. She convenes the Teacher Education AI Network (TEANs).

Foundational Impact is a Good Future Foundation podcast exploring how artificial intelligence is changing education. Through conversations with people whose work is shaping the direction education takes next, we examine the evidence, the challenges facing schools and the decisions being made across the sector.Hosted by Daniel Emmerson.Follow Foundational Impact on Spotify, Apple Podcasts or wherever you listen.Website: www.goodfuture.foundation

Episode OverviewIn this episode from the Festival of Education at Wellington College 2026, host Daniel Emmerson speaks with Daisy Christodoulou, Director of Education at No More Marking. Daisy discusses the role of artificial intelligence in educational assessment, emphasising the importance of actionable feedback over lengthy written notes, the mechanics of comparative judgment, and why handwritten work remains vital for learning. She also addresses the potential pitfalls of AI in schools and offers key guidance for school leaders navigating integration.Key Topics & Timestamps00:00 Speed As Quality00:36 Meet Daisy Christodoulou01:57 Is Written Feedback Useful02:29 Thermostat Feedback Model03:32 Comparative Judgment Explained04:53 AI Judges And Benefits06:09 Avoiding The Doom Loop10:48 Student Trust And Emotions12:48 Fast Feedback Tradeoffs14:59 Handwriting Versus Typing17:14 Digitizing Handwritten Work17:57 Assessment Tradeoffs And Standards20:21 Advice For School Leaders21:56 Hopes For AI In SchoolsKey Takeaways & HighlightsThermostat vs. Thermometer: Assessment feedback shouldn't just measure performance; it needs to change performance by providing actionable steps such as personalised quizzes.Speed as Quality: A scrappy, immediate response can be significantly more effective for student improvement than a polished response delivered weeks late.The 90/10 Rule: Integrating AI into comparative judgment (90% AI, 10% human judging) ensures every paper is reviewed multiple times while maintaining human oversight.Cognitive Differences in Mediums: Research shows moving tests to screens can lower student outcomes compared to paper-based assessments, making handwriting recognition a vital bridge for digital reporting.Action on Shadow AI: Leaders who delay setting AI policies risk teachers resorting to unvetted, consumer-facing tools during high-workload periods.Resources MentionedNo More Marking: Comparative judgment platform for student assessment. https://www.nomoremarking.com/John Jerrim / PISA Research: Natural experiment study comparing student performance on paper vs. digital test formats. https://johnjerrim.com/papers/Guest Contact InformationGuest: Daisy Christodoulou (Director of Education, No More Marking)Daisy Christodoulou is the Director of Education at No More Marking, a provider of comparative judgement assessments. She works closely with schools in the UK, Australia and USA on developing new approaches to assessment, and on applications of artificial intelligence in the classroom.Before that, she was Head of Assessment at Ark Schools, a network of academy schools. She has taught English in two London comprehensives and has been part of UK government commissions on the future of teacher training and assessment.Daisy is the author of three books about education: Seven Myths about Education (2013), Making Good Progress? The future of Assessment for Learning (2017), and Teachers vs Tech (2020).

Our guest in this episode is Dr Fiona Aubrey-Smith, an independent researcher who studies the relationship between pedagogy and digital tools, and the lived experience of students in the classroom.Fiona opens by explaining why identical classroom practices can produce completely different outcomes depending on the belief system behind them. Her example of two teachers greeting students at the classroom door in seemingly the same way, but for very different underlying reasons, shows how policies alone don't determine impact; the intent behind them does.The conversation moves into belonging and dignity, including research showing that struggling students typically ask for help only 1-3 times in a lesson regardless of how much support they need, while confident students almost never ask to be pushed further. Daniel and Fiona discuss the tension between protecting a student's dignity and the value of productive vulnerability.From there, Fiona lays out her framework for introducing technology into schools: start with the purpose of school, define the role of the learner, describe what the learning experience should feel like, define the role of the teacher, and only then decide which tools will help deliver it. Starting with the tool itself, before asking why, is a costly and common mistake.Two school case studies illustrate this in practice. At Haileybury, an embedded one-to-one, on-demand digital provision (available at the point of need rather than simply issued to every student) has freed teachers to rethink classroom layout entirely. At Connect Education Trust, a similar model has extended digital skills and connectivity into the home, empowering families to support learning without needing their own digital capability, coinciding with outstanding inspection outcomes.On generative AI, Fiona shares some thoughtful use cases she sees, such as chaining AI tools to help secondary students revise, and using AI image generation to help primary students with their descriptive writing.She wraps up with a challenge for school leaders. Rather than starting with policies or tools, ask what your school would leave behind if it closed tomorrow, since that gap is where the real human value lies. From there, let AI and digital tools take on the automation and acceleration, so teachers are freed up for the parts of the job only a human can do.

What does AI mean for the way young people relate to each other and to technology itself? That's the focus of the conversation in this episode when Daniel speaks with Jonathan Baggaley, CEO of the PSHE Association and former head of the Child Exploitation and Online Protection Centre.It begins, perhaps unexpectedly, with a Sun Ra poster. Jono uses it to make a point that runs through everything that follows. That is, many of the narratives about AI are being presented as inevitable, and educators have both the right and the responsibility to push back on that.PSHE has been on a long road to statutory recognition in England. Jono maps that journey from the near-miss of the 2010 election wash-up to the introduction of compulsory Relationships, Sex and Health Education in 2019. The results are becoming visible. Between 2022 and last year, the share of young people rating their RSE as good or very good roughly doubled. Yet the subject now faces a technology landscape that changes faster than any curriculum can follow, and it was never designed for that.Much of the conversation focuses on something schools have been slow to address, which is the emotional attachment to AI chatbots. Evidence from a Reddit study found that 60–70% of people who had developed an AI relationship hadn't gone looking for one. They'd simply been chatting. Jono explains this isn't surprising as humans are neurologically primed to anthropomorphise from infancy, attributing intention and agency to moving shapes before they can speak. Even Jono, who understands exactly what large language models are, has caught himself feeling the presence of something in a chatbot conversation. The question is how schools can build resilience to that instinct by giving young people the understanding they need to use them critically.Language, Jono argues, is part of the problem. The word "chatbot" suggests conversation, when what's actually happening is an engagement with a probabilistic automation system. Shifting that framing is one starting point. Another is cross-curricular. Computing can explain the move from rule-based to data-driven AI, citizenship can explore the environmental and geopolitical implications, and PSHE can build on its existing relationships curriculum to examine what young people are and aren't getting from these interactions.The episode sits with a harder question too. Consumer AI is designed for frictionless ease, age verification is often a checkbox, and young people will encounter these tools whether schools introduce them or not. But Jono is clear-eyed about what the evidence shows. When young people genuinely understand what they're dealing with and what they're trading away, they are willing to increase their own friction. None of this requires a new policy or a perfect tool. The conversation is the intervention, and any curriculum that starts with the human already has what it needs.