AI For Learning: What we learned in 2024

To round off the year, we reflect on the themes we have seen from talking to learning organisations about how they are leveraging gen AI.

and

Dec 19, 2024

👋 This week, we’re wrapping up the year by reflecting on one of the big themes, gen AI. We’ll be back on 9th Jan.

Get something in the diary for the new year: our online meetup on 14 January with Ian about Opportunity Solution Tree for Learner Insights from the V&A Academy and our pre-Bett meetup at Zen Educate on 21 January exploring taking bets on new markets. Plus, Marion’s January cohort programme Designing Learning as a Product is open for admissions.

Over the past months, we’ve had the privilege of engaging with a range of forward-thinking organisations using AI to reshape learning.

From established organisations like Oak National Academy, Activate Learning and OpenClassrooms to innovative players like English Coach AI, Apolitical and Junto, each has shared inspiring stories of leveraging AI to tackle real-world challenges.

By reflecting on their approaches, we identified common threads that offer a roadmap for successfully integrating AI into learning. We’ve identified some common use cases:

Tutoring, coaching and role playing
Taking learning from a passive experience to an active one
Making learning more contextual, personal and localised
Helping to broaden access e.g. translation, adjusting the reading age
Making course/lesson development more efficient

But, perhaps more importantly, there are some consistent approaches that are paying dividends. Here’s a closer look at how these leaders are transforming education and the lessons we can build on in 2025.

1. Start with the Problem

Every successful AI initiative we explored identified primarily their challenge and clear outcome, after a period of experimentation with the tools. For these organisations, AI isn’t just a shiny new technology, it’s a tool to help further their existing mission.

At Activate Learning, the problem wasn’t teaching GCSE content; it was helping adult learners overcome self-doubt. Dr Fumiko Pescott explains, “Our learners often struggled with confidence, even when they knew the material. The AI tutors provided a safe, non-judgmental space to practise and grow.”

Similarly, Junto identified a gap in leadership training by helping managers practise tough scenarios, like delivering feedback. "Role plays can be nerve-wracking," says Allen Sanchez, Head of Learning Product. "The AI created a safe space for learners to build confidence before engaging in peer interactions."

An opinion that English Coach AI shares when it comes to language learning role playing. “Students would practise speaking with each other, but it often lacked realism and consistency,” CEO Carla Wyburn recalls from her years as an educator. "We realised AI could offer a consistent, non-judgmental practice partner."

For Apolitical, the challenge was the siloed nature of knowledge in government. Their AI connects civil servants to contextualised resources, transforming scattered information into actionable insights that are relevant for their local context.

Meanwhile, Oak National Academy addressed a more tangible pain point: teacher workload. “We want to give teachers their Sunday nights back,” says John Roberts, Oak’s Director of Product and Engineering. Their AI assistant, Aila, supports teachers in lesson planning, reducing prep time from 50 minutes to just 15.

For OpenClassrooms, the challenge was operational: maintaining and updating a vast course catalogue. “With over 600 courses, mostly on technical topics, the need for updates is constant,” says Senior Learning Designer Laura Besnier. AI helped streamline routine tasks like generating chapter summaries and quizzes, freeing the team to focus on higher-level creativity.

2. Build on Your Unique Advantage

Once the problem is clear, the next step is to leverage existing strengths and resources. One common pattern we learnt: AI is most powerful when it enhances what an organisation already does well and makes use of an existing knowledge base.

Oak National Academy brought its vast repository of National Curriculum-aligned resources to the table, using them as an anchor for its AI outputs. They also codified the expertise they have developed in effective lesson design into a 9,000 word prompt. This approach ensures Aila - their AI-powered assistant - produces content that’s relevant, accurate and trusted by educators.

Apolitical leaned into its eight-year archive of global government discussions, articles and course content to offer public servants highly contextualised learning experiences and join-the-jots between the knowledge that exists in different government silos.

Similarly, Activate Learning embedded its growth-mindset learning philosophy into the behaviour of its AI tutors. “Technology can only be as good as the values it’s built upon,” notes Fumiko. By aligning their AI with their ethos of empowerment, Activate Learning ensured it complemented their broader educational mission.

3. Keep Humans in the Loop

Across every case study, one principle stood out: AI doesn’t replace human expertise, it enhances it. Keeping ‘humans in the loop’ ensures that AI serves as a collaborator rather than a standalone solution.

At Oak National Academy, Aila supports teachers by guiding them through lesson planning but leaves the creative decisions to them. “The teacher remains in control,” explains John. “It feels like you’re planning a lesson, despite the support,” he says. Aila keeps teachers in the driving seat, forcing them to think through the lesson in individual steps, rather than a single process.

For Junto, AI acts as a preparatory step, enabling learners to practise with confidence before engaging in live role-plays with peers. It features as part of a live lesson, led by instructors providing support and accountability.

Activate Learning also found a similar balance, with AI handling repetitive feedback tasks while educators oversee grading. "Feedback is a collaborative effort between AI and humans," says Patrick Kelly-Goss. This synergy ensures that AI remains a trusted tool rather than a shortcut that compromises quality. Whilst at the same time reducing marking workloads by 65%.

A similar approach is taken by OpenClassrooms in their learning design process which uses AI as a “technical reviewer” to help subject matter experts refine course content. “It points out inconsistencies or missing elements, but final decisions still lie with the humans,” says Laura.

4. Focus on Adding Value

The true potential of AI lies in its ability to add value where it matters most, rather than reducing costs. For teachers, this might create more time for valuable tasks or improving the quality of the materials; for learners, it’s about making learning more contextual, active and building their confidence and clarity.

Oak National Academy’s Aila has been a game-changer, saving teachers up to 3–4 hours per week on lesson planning. "There's no point reinventing the wheel," John says. “We help teachers adapt high-quality lessons to meet their specific needs.” It also frees them up to spend time where they can add most value: working with pupils in the classroom.

At Activate Learning, value comes from the personalised support AI tutors provide. By breaking tasks into manageable steps, these tools help learners develop confidence in their abilities. As Fumiko summarises, “it’s about fostering independence while maintaining a supportive environment.”

English Coach AI adds value by contextualising learning, offering visuals and scenarios specific to learners’ environments as well as an “Ask me Anything” feature which focuses on providing immediate, actionable real-time clarification to learners.“The ability to personalise content makes a huge difference in engagement,” says Carla.

5. Experiment and Co-Create

Innovation thrives on experimentation and collaboration. Each organisation we spoke with embraced a culture of testing, learning, and refining alongside their users and released experiments early.

Oak National Academy took an open-source approach, sharing Aila’s development with the sector to encourage feedback and improvements as well as involving educators in a closed and then open beta.

Activate Learning brought students into the design process, creating a “sandbox” for experimentation. “We didn’t just look at metrics; we spoke with learners and educators to build trust and refine the tools,” explains Patrick.

For Apolitical, starting simple with a “vanilla” course assistant, with minimal user interface or instructions, allowed them to gather insights on user behaviour and refine their design based on real-world interactions.

Meanwhile, Junto fast-tracked their AI leadership coach prototype in just three days by focusing on a specific part of the learner journey, enabling them to test and iterate rapidly. As Allen notes, “Move fast, but measure what matters.”

Looking ahead to 2025

One common denominator: Integrating AI into education isn’t just about deploying new technologies.

It’s about empowering people—learners, educators, and organisations—to reach their full potential.

By starting with the problem, leveraging strengths, keeping humans in control, focusing on adding value, and embracing collaboration, these organisations offer a blueprint for meaningful innovation.

“It’s not just about knowledge transfer anymore” reflects Lowell from Apolitical. “It’s about helping learners apply their skills to their context and challenges.” This focus on application and personalisation is shaping the future of education—one thoughtful, transformative step at a time.

As 2024 closes, it’s clear that gen AI is starting to move beyond novelty and bolt ons. It’s a catalyst for reimagining how we teach, learn, and connect.

For organisations willing to experiment, adapt, and collaborate, the possibilities are limitless. The question is no longer if AI will transform education but how we can harness its potential wisely - and mitigate its risks - to create a better, more inclusive learning experience for all.

Find out more about each of these examples by following the links on our AI for Learnin g series. Share your examples in the comments - we’re looking for more case studies to write about next year.