top of page
Noodle-factory-logo
Book a Demo
Get Started

Rethinking Learning with AI | A webinar series with the University of London. Save your spot →

What Vocational and Polytechnic Students Need from AI (That University Students Don't)

Marielyn Wong
11 minutes ago
8 min read

The instructor is standing in a workshop in Bandung. Around her, seventeen students are halfway through stripping down a diesel engine, hands greasy, manuals open, concentration written on their faces. On a whiteboard behind her, someone has scrawled a note about a YouTube video that was supposed to demonstrate the torque sequence for the cylinder head. The video buffered and then gave up.


She has tried one of the popular AI chat tools her polytechnic recently subscribed to. It is, she says, designed for essay writing. She could ask it to explain torque in general terms. She cannot ask it to verify whether a student's hands are positioned correctly, or to flag that the student has missed a step in the sequence that will matter badly when the engine runs again.


This is the gap that most AI-in-education coverage does not see, because most AI-in-education coverage assumes a lecture hall.

The Invisible Majority


Globally, vocational and technical education serves more students than most people realise. In Southeast Asia alone, the scale is striking: Indonesia's SMK (Sekolah Menengah Kejuruan) vocational high school network enrols more than 5 million students. Thailand's TVET system covers roughly 1.1 million. Singapore's Institute of Technical Education and five polytechnics together graduate around 25,000 students annually, in programmes ranging from aerospace engineering to pastry arts to marine technology.


These students do not learn the way a law or economics undergraduate learns. They learn by doing. Their assessments are not primarily essays and examinations; they are competency demonstrations, skills checkoffs, workplace practicums, and industry certifications. And when they fail, the consequences are not a lower grade. They are a faulty weld, a mis-dosed medication, a building that doesn't meet code.


The AI tools being built for education, by and large, were not designed for them.

What AI Was Built For (And What It Wasn't)


The dominant model for AI in higher education has converged on a few core use cases: essay feedback, reading comprehension support, exam preparation, and research assistance. These are genuinely useful, and the market has responded accordingly. Tools like Grammarly, Turnitin's AI capabilities, and a wave of AI tutoring platforms are optimised for text-heavy, knowledge-recall learning.


Vocational education is different in almost every structural dimension. Its knowledge is embodied: in muscle memory, in the feel of a correct weld, in the smell of a dough that has proofed too long. Its assessment criteria are competency frameworks, not rubrics for analytical writing. Its standards are often set not by institutions but by industry bodies: the ASEAN Skills Competition, the City & Guilds framework, the Workforce Skills Qualifications in Singapore.


Ask a general AI tutor to explain how to perform a four-point lift on a vehicle, and it will give you a reasonably accurate text description. Ask it to assess whether a student has performed the lift correctly, and it has nothing to offer. The gap between what these tools do well and what vocational education actually needs is not a minor product gap. It is a design mismatch.

Three Distinct Needs Vocational AI Must Address


First: practical skills verification. The challenge of using AI to assess hands-on competency is real and unsolved at scale. Some promising approaches are emerging: computer vision tools that can analyse video of a student's technique in real time, haptic simulation systems used in surgical training, and AR overlays in manufacturing that flag procedural deviations. Singapore Polytechnic has piloted AI-assisted assessment in some engineering modules, using video analysis to check procedural sequences.


But these tools are expensive, technically demanding to deploy, and not yet widely available. For most vocational institutions across Southeast Asia, the more immediately actionable question is: where can AI reliably help, right now, without requiring robotics labs or AR infrastructure?


The answer is in the surrounding knowledge: the theory, safety procedures, industry standards, and troubleshooting logic that underlies every practical skill. A student who knows why the torque sequence matters, what goes wrong when it is skipped, and how to diagnose the resulting fault is a better technician. AI can teach that knowledge, verify that the student can articulate it, and do so at scale.


Second: industry-standard alignment. Vocational curricula are not generic. A culinary programme at Temasek Polytechnic follows different standards from a culinary programme at a Bangkok hospitality institute. An electrical installation course at an ITE must align with Singapore's SS638 wiring standard. A nursing aide programme in Indonesia must map to national KKNI competency levels.


General-purpose AI cannot navigate these distinctions unless it has been specifically configured to know them. Asking ChatGPT to quiz a student on Singapore electrical standards will produce something plausible but unvetted. What vocational institutions need is AI that knows their specific competency frameworks, that can ask questions keyed to the exact units of competency their assessors will evaluate, and that can flag when a student's understanding of a procedure doesn't match the standard the industry actually requires.


Third: multimodal and practical content support. Vocational learners often struggle with dense written material, not because of low ability, but because their learning style is practical and their entry qualifications are not always text-heavy. AI that can explain a concept multiple ways, generate worked examples using equipment the student has actually handled, or produce a procedural walkthrough in plain language can dramatically lower the barrier to theory comprehension. This is a genuine win that does not require simulation technology. It requires content-aware AI that knows what the student is training to do.

What Institution-Specific AI Actually Looks Like


The practical question, then, is not whether AI can serve vocational education. It is whether AI can be configured specifically enough to be useful.


Noodle Factory was built around a specific conviction: that AI in education only works when it is grounded in the educator's context, not a generic knowledge base. For vocational education, where the knowledge is not just academic but procedural, industry-specific, and tied to particular competency frameworks, that principle is not a design preference. It is a prerequisite.


The platform's AI tutor, Walter, grounds itself entirely in the institution's own course materials — theory notes, safety procedures, competency frameworks, and assessment criteria. A student asking Walter about diesel engine maintenance is not getting a Wikipedia summary of how engines work. She is getting answers drawn from the exact content her assessors will test her on. That alignment matters more in vocational contexts than anywhere else, because the cost of a gap between what a student learned and what the industry actually requires is not a failed essay. It is a safety incident.


But knowledge recall is only part of what vocational learners need. The other part is practice — and this is where the platform's Role Plays feature addresses something genuinely hard to replicate in a classroom. Role Plays lets educators design scenario-based sessions where students face realistic, industry-specific situations in real time: a customer complaint in a hospitality programme, a safety incident response in an engineering module, a patient consultation in nursing. Students do not just read about what to do. They practise doing it, receive feedback, and try again. It is as close to simulation as text-based AI can get, and it works within the educator's defined scenarios and rubrics — not a generic script the vendor wrote.


For institutions that need to go further, Custom Skills allows a programme to wire AI capability directly to its specific competency frameworks, assessment rubrics, and industry standards. A polytechnic running a precision engineering programme can configure the AI to the specific tolerances, procedures, and grading criteria that its students will be evaluated against. This is how institution-specific knowledge gets operationalised, rather than relying on the AI to interpret a general-purpose competency description and hope the interpretation matches.


And because vocational students learn by explaining and doing, not just by receiving, NoodleMe lets students build their own AI tutors from their notes and share them with classmates. This is particularly resonant in vocational contexts, where peer learning and practical knowledge-sharing are already embedded in how education happens — on the workshop floor, in the kitchen, in the clinic.


Together, these tools reflect a platform designed around the idea that different parts of a learning journey have different needs. The theory has to be grounded in the right standards. The practice has to simulate the right scenarios. The assessment has to reflect the right criteria. And the learner has to be able to engage with all of it in a way that matches how she actually learns.


This is what institution-specific AI looks like in a practical-skills context: an AI that knows the difference between the competency standard for a beginning electrical trainee and an advanced one, that can put a nursing student through a patient consultation scenario and give her feedback against clinical rubrics, and that can escalate to a human instructor when a student's question moves from the theoretical into the hands-on, where human judgment is irreplaceable.


The goal is not to replace the instructor in the workshop. The goal is to make sure that when students are in the workshop, the conceptual knowledge that underpins their technique is solid enough that the instructor can focus on what only a human can do: watch the hands, read the hesitation, and adjust the teaching in real time.

The Conversation Southeast Asia Needs to Have


The edtech sector's fixation on university settings is not just a marketing blind spot. It has policy consequences. When ministries and institutions evaluate AI tools for education, the conversation tends to default to essay grading, plagiarism detection, and academic research support, because those are the use cases that have the most visible, vocal advocates.


Vocational educators, who are in much of Southeast Asia training the workforce that will actually build, maintain, and operate the economy, are largely absent from that conversation. They are tired of receiving advice that assumes they have a reading list and an exam schedule. Most of them are working with competency checklists, industry mentors, and a persistent shortage of equipment and instructors.


UNESCO's 2023 Global Education Monitor flagged TVET as chronically underfunded relative to its enrolment and economic importance, noting that vocational systems consistently receive less government and private investment per student than academic pathways. AI investment in education is at risk of replicating that disparity if the tools being built and promoted continue to default to the university model.


The Bandung instructor with the diesel engine and the buffering video is not an edge case. She is the centre of the story. AI that is built around her students' actual learning tasks, not retrofitted from a different context, will do more for educational equity in Southeast Asia than another essay-feedback tool.


That is the conversation worth having. It has barely started.

Further Reading


Sources


bottom of page
Chat with Walter