Reflecting on Sprint 1: Exploring the role of AI in supporting self-paced TVET learning
Applying a sandbox approach to SEA-VET Learning
As part of the ASEAN–UK SAGE programme, SEAMEO VOCTECH has partnered with EdTech Hub to leverage the Hub’s sandbox approach and methodology to strengthen SEA-VET Learning, the online learning component of the SEA-VET.net platform. SEA-VET Learning is designed to play an important role in expanding access to vocational education and training across Southeast Asia, particularly through self-paced and asynchronous courses. However, like many online learning platforms, sustaining learner engagement and supporting course completion remains a persistent challenge.
Anchored in a Do–Measure–Learn cycle, sandboxes create a ‘safe space’ to test assumptions, generate rapid evidence, and iterate before committing to larger-scale changes. Rather than aiming to implement full solutions upfront, the focus is on answering the most critical questions at the right time, so partners can make better-informed decisions about what to scale, adapt, or deprioritise.
The SEAMEO VOCTECH sandbox is one of several sandboxes we’re conducting with regional education partners and is deliberately scoped to complement — rather than duplicate — SEAMEO VOCTECH’s ongoing platform and sustainability initiatives.
Read more about the wider programme and this series of blogs here, as we unpack how we are learning by doing: testing ideas for Southeast Asian education challenges through real-world experiments, be it in the realm of policy or programme design.
Sprint 0: Grounding the sandbox – the sprint before the sprints
Sprint 0 focused on building a shared understanding of SEA-VET Learning and aligning the sandbox with VOCTECH’s broader priorities. Through analysis of backend platform data, a review of internal strategy documents, and close engagement with the SEAMEO VOCTECH team, we examined how learners currently interact with the platform, which courses attract the most engagement, and how success is being measured.
Sprint 0 successfully generated important process learnings. Early and open coordination with SEAMEO VOCTECH revealed that significant diagnostic work had already been conducted earlier in the year, including internal workshops that identified gaps and potential strategies for improving engagement. With this information, we were able to adapt our sprint process to exclude a repeat of the same brainstorming activities to co-develop a theory of change with the SEAMEO VOCTECH team, and move to understanding and evaluating identified goals, targets, solutions, and testable assumptions.
Initial insights from these workshops and our own discussions with the team supported potential solutions such as progress-tracking indicators and micro-credentialling to drive greater engagement and retention. As SEAMEO VOCTECH progressed parallel work on platform sustainability with other partners to potentially enable micro-credentialing, the sandbox scope was further refined to focus explicitly on learner-facing features within self-paced courses, thus enabling a key feature of the sandbox approach: to modify, refine, and evolve as information surfaces and findings are made.
Keeping consistent with that approach, three key lessons from Sprint 0 shaped the design of Sprint 1:
- Coordination matters – early alignment helped avoid duplication and ensured the sandbox added distinct value.
- Metrics need context – conventional engagement metrics did not always reflect how learners actually use the platform, particularly where content is downloaded for offline use.
- Focus enables expediency – narrowing the scope made it possible to move quickly from discovery to experimentation.
Sprint 1: Exploring whether AI can support self-paced learning
As insights from Sprint 0 set the context and constraints and clarified already identified and discussed solutions, it allowed us to manoeuvre Sprint 1 into investigating an untried potential solution: the potential use of an AI chatbot that can support self-paced and asynchronous learning.
Why AI, and why now?
Insights from Sprint 0 pointed to a central challenge: while VOCTECH wanted to increase SEA-VET Learning’s self-paced and asynchronous offerings, learners consistently value instructor presence, feedback, and peer support — elements that are often missing in self-paced formats. This led to a key question for Sprint 1:
Could an AI-powered chatbot meaningfully support self-paced learning on SEA-VET Learning by replicating some aspects of instructor support?
Rather than assuming AI was the right solution, Sprint 1 was designed to test the utility and desirability of an embedded AI chatbot, and to understand how learners might realistically use it to develop well-grounded and evidence-backed use cases.
Sprint 1 approach and activities
Sprint 1 focused on gathering learner perspectives across the region to assess interest in an AI chatbot and identify priority use cases.
The sprint tested three core assumptions:
- SEA-VET Learning users would use an AI chatbot if it were embedded on the platform.
- Instructor support is a key driver of engagement in hybrid and online courses.
- An AI chatbot could replicate some aspects of instructor support in self-paced learning.
To gather data to answer the key question and test the core assumptions, the following activities were conducted:
- A comprehensive survey was distributed to participants of previous SEA-VET regional trainings to inquire about their current usage of AI and to break down and investigate motivating factors for continued learning and course completion.
- A shorter survey was disseminated via social media and messaging platforms to bolster initial survey findings and to leverage the fake door method to gauge interest in a chatbot by providing a potential doorway to early access to the chatbot.
- Follow-up interviews with the interested learners were carried out to explore themes emerging from the survey data in more depth.
Key findings from Sprint 1
Sprint 1 findings helped clarify both the potential and the limits of AI for self-paced TVET learning.
Learners are already using AI — but selectivelyA large proportion of respondents reported already using tools like ChatGPT to support their learning, particularly for quick questions and clarification. This suggested a relatively low barrier to adoption for an embedded AI chatbot, provided it addressed real needs rather than introducing new complexity. |
AI is most valued as a support tool, not a motivator
Learners expressed the strongest interest in AI use cases that reduce friction, such as:
- Answering questions about course content
- Providing translated explanations
- Suggesting next steps or additional resources.
Career guidance is promising, but expectations vary
Responses suggested moderate interest in AI-supported career guidance, with similar levels of preference for on-demand information and more interactive guidance. This pointed to potential value but also highlighted the need for careful design to manage expectations.
Learners showed interest and experience in leveraging AI chatbots (such as ChatGPT and Claude), but by contrast, there was much lower interest in AI-driven reminders or motivational nudges. Regular feedback from instructors and peer interaction remained the most valued forms of support for staying engaged.
Interviewed TVET instructors did express some nuanced concerns around the use of AI. They had thoughtful apprehensions about AI reliability, over-reliance, and the need for human and expert oversight. Importantly, data privacy did not emerge as a major concern, but there was a clear need for AI literacy and guardrails to ensure AI supports — rather than replaces — learning.
Identified potential use cases for SEA-VET Learning
Lowering access barriers through multilingual and navigational support
- An AI chatbot could support learners across the region by providing bilingual or mixed-language responses, helping with platform navigation, and offering explanations of technical concepts in learners’ preferred languages.
Supporting career guidance and next-step decision-making
- The chatbot could provide career counselling by asking guided questions, responding to learner-defined goals, and offering information on relevant learning pathways and in-demand skills aligned with labour market needs.
Strengthening learning through prerequisite and just-in-time academic support
- AI could help learners identify prerequisite knowledge gaps, provide targeted explanations or diagnostics, and support revision and recall aligned with course content, particularly in self-paced settings.
From Sprint 1 to Sprint 2: Shifting from desirability to feasibility
Taken together, Sprint 1 findings reframed the role of AI in SEA-VET Learning. Rather than acting as a motivation engine, AI appears most valuable as a contextual support layer that helps learners navigate content, overcome language barriers, and access timely explanations in self-paced courses.
These insights shape the focus of Sprint 2. The next sprint will move beyond desirability to assess feasibility, examining:
- Technical and operational requirements
- Cost implications
- Inclusivity and multilingual considerations
- Data protection and maintenance needs
Sprint 2 will support VOCTECH to weigh different AI integration options using an evidence-informed decision matrix, ensuring any future investment in AI is both strategic and sustainable.
What’s next?
The VOCTECH sandbox demonstrates how a sandbox approach can support better decision-making — not by rushing to implement new technologies, but by generating timely evidence, aligning with partner priorities, and remaining responsive to learning along the way. As the sandbox moves into Sprint 2, the focus remains on helping VOCTECH decide whether, where, and how AI can responsibly support self-paced TVET learning across Southeast Asia.
