AI for Asesmen Nasional Preparation in Indonesia

The Asesmen Nasional rewards reasoning, not recall, so AI's most useful role in preparing for it is generating reading and numeracy items that force students to interpret and apply — not drill facts the way the old Ujian Nasional once did. Used well, AI can build practice sets modeled on AKM's format in minutes; used carelessly, it just reproduces the memorization habits the assessment was designed to move schools away from.

Quick Answer: AI can help Indonesian teachers build AKM-style literacy and numeracy practice items, character-survey discussion prompts, and computer-based-test familiarization exercises. It can't replace understanding the Asesmen Nasional's actual purpose — diagnosing school and system quality through a student sample, not ranking individual students or determining graduation — which should shape how any AI-generated material gets used in class.

Indonesia replaced its old national exit exam with a fundamentally different kind of assessment in 2021, and that shift changes what "exam prep" itself should look like. For how this compares with assessment reform elsewhere, see AI in Education Around the World: A 2026 Regional Guide.

What the Asesmen Nasional Actually Tests

The Asesmen Nasional (AN), introduced by Indonesia's education ministry in 2021 and now administered under Kemendikdasmen (Kementerian Pendidikan Dasar dan Menengah, formed in a 2024 ministry reorganization), replaced the older Ujian Nasional entirely. Unlike its predecessor, AN doesn't determine whether an individual student graduates.

Because AN samples students rather than testing every student, and reports results at the school and system level, no individual student passes or fails it. That single design choice should reshape how "AN prep" gets framed in a classroom — the goal is building the underlying literacy and numeracy skill, not coaching specific students to clear a personal threshold.

Why AN Prep Differs From Drilling for the Old UN

The Ujian Nasional rewarded memorized subject content tied closely to a textbook. AKM deliberately does not — it's built to measure whether a student can reason with unfamiliar material, which means practice questions that just recycle textbook facts miss the point entirely.

AKM Literacy: Reasoning Over Recall

AKM Literasi presents unfamiliar passages — informational and literary — and asks students to interpret, evaluate, and connect ideas across them, not recall a fact stated directly in a textbook. A strong practice item should require inference, not memory.

AKM Numeracy: Applied Problem-Solving

AKM Numerasi spans four content domains (bilangan/number, aljabar/algebra, geometri/geometry, and data dan ketidakpastian/data and uncertainty) crossed with cognitive levels from basic understanding through application to full reasoning. Real-world scenario framing — a market price, a travel schedule, a measurement problem — is central to the format, not an occasional flourish.

• Avoid: "What is 45% of 200?"
• Closer to AKM's actual style: "A shop discounts an item by 45%. If the discounted price is Rp 110,000, what was the original price, and how do you know?"

The Survei Karakter and the Pancasila Student Profile

AKM gets most of the "AI prep" attention, but the Survei Karakter measures something entirely different: how well a school's culture reflects the Pancasila Student Profile (Profil Pelajar Pancasila), the six-dimension character framework underlying Indonesia's current curriculum policy.

• Beriman, bertakwa kepada Tuhan Yang Maha Esa, dan berakhlak mulia — faith, piety, and noble character.
• Berkebinekaan global — global diversity awareness.
• Bergotong royong — mutual cooperation.
• Mandiri — independence.
• Bernalar kritis — critical reasoning.
• Kreatif — creativity.

There's no "correct answer" to drill for here, since the survey captures perception and school climate rather than testable knowledge. Where AI genuinely helps is generating classroom discussion prompts and reflective journal questions built around each dimension, giving students regular, low-stakes practice articulating how these values show up in their own schoolwork and behavior — not preparing them to answer a survey a particular way.

Grade-Band Differences: Preparing Grade 5, 8, and 11 Differently

AN samples students at three different points — Grade 5, Grade 8, and Grade 11 — and treating all three identically wastes an AI tool's biggest advantage: producing genuinely grade-appropriate material fast.

For teachers covering Grade 5 through Grade 9 specifically, the practical takeaway is that a single AKM-style prompt template rarely transfers cleanly between a Grade 5 and a Grade 8 class — regenerate with the grade band named explicitly rather than reusing one set across both.

What Happens to the Results: Rapor Pendidikan

Because AN reports at the school level, results feed into Rapor Pendidikan (the Education Report Card), a government dashboard schools use for improvement planning rather than a score handed back to individual students or parents. Understanding this changes what "using AN data" should look like for a teacher or school leader.

• A school's Rapor Pendidikan combines AKM results with the Survei Karakter and Survei Lingkungan Belajar into indicators meant to guide a school improvement plan, not rank the school publicly against others.
• AI can help draft an improvement-plan narrative from a school's own Rapor Pendidikan indicators — summarizing patterns, suggesting focus areas — but the underlying data and any resulting decisions still need a school leader's direct review.
• Teachers benefit most from knowing which literacy or numeracy sub-skills a school's own data flags as weak, then targeting AI-generated practice at exactly those sub-skills instead of AKM prep in general.

How AI Can Help Build AKM-Style Practice Materials

Generating AKM-aligned items is less about the AI tool and more about specifying the reasoning demand explicitly, since a generic "make practice questions" prompt defaults to recall-style items.

1. Specify the AKM domain directly — literacy text type, or numeracy content domain and cognitive level — rather than a general subject name.
2. Ask for an unfamiliar scenario, not a textbook-recycled one, since AKM's whole design tests transfer to new material.
3. Request a mix of cognitive levels in one set: some items at basic understanding, most at application and reasoning, mirroring AKM's actual distribution.
4. Ask for open-response items alongside multiple-choice ones, since AKM includes short-answer and essay-style items that reward explaining reasoning, not just selecting an answer.
5. Generate an answer key that explains the reasoning path, not just the correct option — this doubles as a model for how students should show their own thinking.
6. Batch a full set across both literacy and numeracy in one planning session, then review the whole set for genuine reasoning demand before using any of it.

A Sample Prompt Walkthrough

A side-by-side comparison shows how much a prompt's specificity changes the output's usefulness for AKM prep specifically.

• Weak prompt: "Give me 5 numeracy questions for Grade 8."
• AKM-aligned prompt: "Generate 5 AKM Numerasi-style questions for Grade 8, covering the data-and-uncertainty domain, each set in an everyday Indonesian context (market prices, travel time, or school data), with at least 3 requiring application-level reasoning rather than direct calculation. Include a written explanation of the reasoning for each answer, not just the final number."

The weak prompt will likely return standard textbook-style arithmetic. The AKM-aligned prompt forces context, domain, and cognitive-level specificity — the three things that actually distinguish AKM items from ordinary practice questions.

The same gap shows up on the literacy side, where the fix is asking for an unfamiliar passage plus interpretation questions rather than direct-recall ones.

• Weak prompt: "Give me a reading passage and 3 questions for Grade 8."
• AKM-aligned prompt: "Write an unfamiliar 200-word informational passage about a local environmental topic, suitable for Grade 8. Then write 3 comprehension questions: one asking students to identify the main idea, one asking them to infer a cause-and-effect relationship not stated directly, and one asking them to evaluate whether a claim in the passage is fully supported by the evidence given."

Notice what the second version does differently: it names a text type, a length, and — critically — the specific reasoning move each question demands, rather than leaving "comprehension question" to default to a recall check.

Building a Term-Long AKM Practice Rhythm

A single practice set the week before testing does far less than the same effort spread across a term, and AI's real advantage here is making frequent, small-batch generation cheap enough to sustain.

1. Map the term's weeks against both AKM domains — literacy and numeracy — so neither gets crowded out close to a deadline.
2. Generate a small set weekly rather than a large set once, so items stay varied instead of repeating the same handful of contexts.
3. Rotate which cognitive level gets emphasis — a week weighted toward basic understanding, the next toward application, another toward full reasoning.
4. Log which sub-skills students consistently struggle with, then feed that pattern back into the next week's prompt.
5. Reserve the final two weeks for full-length, timed, on-screen mock sets rather than new content, since familiarity with pacing matters as much as content by that point.

Computer-Based Delivery: Building Digital Familiarity

AN is delivered through ANBK (Asesmen Nasional Berbasis Komputer), a fully computer-based format, which means comfort navigating an on-screen test interface is itself part of readiness — separate from literacy and numeracy skill.

• Practice on-screen, not only on paper, especially for students with limited home computer access, since navigating tabs, scroll, and on-screen answer selection is its own skill.
• Simulate time pressure occasionally, since pacing across an on-screen interface feels different from a paper test students may be more used to.
• Check school computer-lab capacity early in the term, not the week before testing, since ANBK scheduling depends on available devices and connectivity.

When a School Doesn't Have Enough Devices

Schools without sufficient computers or stable connectivity can administer ANBK by sending students to a nearby "resource" school (sekolah pelaksana) that hosts the test on their behalf, rather than every school needing its own full lab. If your school falls into this category, on-screen practice sessions matter even more, since students may be sitting the actual test on unfamiliar hardware in an unfamiliar room.

Tools That Can Help

EduGenius can generate literacy passages, numeracy word problems, and answer keys with explanations from a saved class profile — useful for building an AKM-style practice bank without starting from a blank page each time, and its content design draws on Bloom's Taxonomy for cognitive-level variety, which maps reasonably well onto AKM's own tiered cognitive demands.

• Platform Merdeka Mengajar (PMM), the Indonesian government's own free teacher platform launched under the Merdeka Belajar policy umbrella, hosts official AKM-aligned training materials and example items worth checking any AI-generated set against.
• General chatbots (ChatGPT, Gemini, Claude) handle open-ended item generation well but need the AKM domain and cognitive level spelled out explicitly, the same as any other tool.
• For a broader look at how Indonesia's competency-linked curriculum shapes lesson planning specifically, see AI Lesson Plans Aligned to CBC/CBE Competency-Based Learning, which covers a comparable shift in several other countries.

Where International Benchmarks Fit In

Indonesia's participation in the OECD's Programme for International Student Assessment (PISA) shares real conceptual overlap with AKM — both measure applied reasoning rather than curriculum recall, and Indonesia's PISA results have consistently trailed the OECD average across reading, mathematics, and science. The World Bank's "learning poverty" framework, which measures whether children can read and understand an age-appropriate text by age 10, frames the same underlying problem AKM Literasi is built to surface at the school level.

That overlap is useful context, not a reason to import PISA-style test prep directly — AKM has its own format, domains, and reporting structure specific to Indonesia's curriculum and the Pancasila Student Profile.

• The pattern is global, not unique to Indonesia: UNESCO's reporting on foundational learning treats reading and math proficiency by age 10 as the base layer every other subject depends on.
• AKM's applied-reasoning design is deliberate, not incidental — it reflects the same shift toward measuring transferable skill, rather than syllabus recall, reshaping large-scale assessment in several countries at once.

Related Reading

• AI in Education Around the World: A 2026 Regional Guide — pillar
• AI for ECAT and Engineering Entry Tests — hub
• AI Lesson Plans Aligned to CBC/CBE Competency-Based Learning
• AI for CBSE Board Exams Preparation in India
• AI Homework Help for Parents in Pakistan
• Best AI for Math Problems in 2026 (Benchmarked)

References

• Kementerian Pendidikan Dasar dan Menengah (Kemendikdasmen) — Asesmen Nasional structure and AKM domains.
• Platform Merdeka Mengajar (PMM) — official teacher training materials and example AKM items.
• OECD — Programme for International Student Assessment (PISA), Indonesia participation and results.
• World Bank — Learning Poverty measurement framework.
• UNESCO — foundational literacy and numeracy reporting.