Learning Exercise
From Data to Action: Explore AI Agents and Decide When a Human Must Approve (Daripada Data kepada Tindakan)
A self-paced exercise (about 60 minutes) for beginners using the Malay-language iOER Asas AI Agentik: Daripada Data kepada Tindakan. Learners work through the five short units, try the lab simulation of a class-room-change announcement, and finish with the self-check quiz and a short reflection. The goal is to tell an AI agent apart from a fixed workflow, see how data, tools and multiple agents work together, and decide when an action must wait for human approval. No programming is needed.
An inclusive, open learning module in Bahasa Melayu (Malay) that introduces agentic AI to beginners without programming.... see more
Exercise
- Start (5 min). Open the resource and read Mula di sini (Start here). Choose how you want to learn in each unit with the Belajar Cara Anda tabs: text, captioned video, slides, infographic, audio or full notes.
- Units 1–2 (15 min). In Kenali ejen AI, compare a generative response, a fixed workflow and an LLM-based agent. In Apa yang ada dalam ejen?, match each component (goal, context, memory, tools, planning, review) to its job in a study task. Complete each unit's Semak kefahaman activity.
- Units 3–5 (20 min). Classify the four kinds of operational data and judge whether the context is enough (Unit 3); follow how agents hand work to each other through events such as
draf_siapandpembetulan_diperlukan(Unit 4); choose the right level of human review for actions with different impact (Unit 5). - Lab simulation (10 min). Open Makmal simulasi. Run the class-room-change case several times with different data conditions (complete data, missing room, conflicting sources, lecturer rejects the draft). Note when the agent may prepare a draft, when it must stop and ask, and who gives final approval. The simulation uses fixed rules and fictional data; no message is sent.
- Quiz and reflection (10 min). Take Uji kefahaman and read the explanation for each answer. Then write a short reflection in Refleksi & sumbang: name one task in your own course or work that should stay a fixed workflow, and one that could use an agent with a human approval step.
Tip for instructors: assign steps 1–3 before class and use step 4 as an in-class discussion.
Disciplines
Technical Notes
Runs in any modern browser on desktop or mobile. Videos and audio include captions and downloadable transcripts; slides are available as PDF. Progress and notes are stored only in the learner's own browser. Accessibility: WAVE 0 errors (AIM 10/10).
Requirements
Basic digital literacy and a web browser. The resource is in Bahasa Melayu (Malay); browser translation can help non-Malay readers, but key terms are best read in the original. No sign-in and no programming are required.
Topics
Agentic AI; AI agents vs. fixed workflows; LLM-based agents; agent components (goals, context, memory, tools); operational data and context; multi-agent collaboration and event-driven design; human-in-the-loop approval; responsible AI action.
Type of Task
Learning Objectives
- Distinguish an AI agent from a generative response and a fixed workflow.
- Explain how data, tools and cooperating agents contribute to an agent's actions.
- Decide which actions need human review or approval, based on their impact.
Assessment
Formative: each unit's Semak kefahaman activity and the 8-question Uji kefahaman quiz, both with explanations. Instructors may collect the short written reflection (step 5) and assess whether the learner justified when human approval is needed.