Building Job-Based AI Assistants: Why AI Needs to Know Who You Are

A laptop screen showing a custom GPT builder configuration for a job-based AI assistant, with a notebook and coffee on the desk

Bikin GPT Berdasarkan Jabatan: Kenapa AI Harus Tahu Lo Itu Siapa

Jadi gini. Kalau lo mau bikin GPT atau AI assistant yang jawabannya beneran berguna buat pekerjaan, kuncinya bukan di pertanyaan yang canggih. Kuncinya di konteks. AI harus tahu lo itu siapa, jabatannya apa, wewenangnya sampe mana, dan lo kerja di lingkungan organisasi yang kayak gimana.

Tapi tunggu dulu. Sebelum gw jelasin lebih jauh soal framework-nya, gw mau cerita dulu kenapa gw sampe kepikiran bikin hal kayak gini.

Gw punya pengalaman yang menurut gw lucu sekaligus nyebelin.

Jadi waktu itu gw lagi coba pakai ChatGPT buat bantu mikirin satu masalah. Anggap aja gw ini seorang Manager Keuangan. Ada masalah di bagian pembayaran vendor. Tagihan numpuk, tapi cash flow lagi ketat. Gw tanya ke AI, kurang lebih gini: “Ada masalah nih, pembayaran vendor telat terus, gimana ya solusinya?”

Terus AI jawab apa coba?

“Sebaiknya koordinasikan dengan Manager Keuangan.”

Wadaw.

Gw sendiri Manager Keuangan bray. Masa gw disuruh koordinasi sama diri gw sendiri?

Nah dari situ gw sadar. Masalahnya bukan di AI-nya bodoh. Masalahnya AI gak tahu gw itu siapa. Dia gak tahu gw jabatannya apa. Dia gak tahu wewenang gw sampe mana. Dia gak tahu gw punya atasan siapa, bawahan siapa, dan keputusan apa yang boleh gw ambil sendiri.

Dia cuma tahu gw nanya soal vendor.

Ya wajar jawabannya generik. Jawaban hasil googling. Jawaban template. Jawaban yang bunyinya kayak artikel wikiHow.

Kenapa ChatGPT Jawabannya Sering Generik Banget?

Ini pertanyaan yang sering banget muncul. Dan jawabannya sebenernya sederhana.

AI itu cuma bisa kerja pakai apa yang lo kasih. Kalau lo cuma kasih pertanyaan, ya dia jawab pertanyaan. Tapi kalau lo kasih konteks lengkap tentang posisi lo, dia bisa jawab sebagai partner berpikir yang paham situasi lo.

Coba bayangin lo punya asisten baru di kantor. Hari pertama dia masuk, lo langsung tanya: “Menurut kamu, gimana cara ngatasin masalah vendor?”

Apa dia bisa jawab dengan baik? Gak bisa. Dia belum tahu lo siapa, departemennya apa, perusahaan lo bidang apa, aturannya gimana.

Tapi kalau lo kenalin dulu. “Nih, gw Manager Keuangan di perusahaan distribusi, gw punya 3 staf, wewenang gw sampe approval 50 juta, di atas itu harus ke Direktur, masalahnya sekarang cash flow ketat.”

Baru dia bisa kasih jawaban yang relevan.

Nah, AI juga sama.

Apa Itu AI Assistant Berbasis Jabatan?

Jadi gagasannya gini. Daripada lo pakai AI generik yang gak tahu apa-apa tentang lo, lo bikin AI assistant yang udah dikasih “job description lengkap” tentang posisi lo.

Misalnya:

• AI Assistant — Manager Keuangan
• AI Assistant — Manager HR
• AI Assistant — Manager IT
• AI Assistant — Manager Operasional
• AI Assistant — Direktur

Dan seterusnya.

GPT-nya gak dimaksudkan buat ngambil keputusan final. Bukan buat gantiin lo. Tapi buat jadi teman mikir. Decision support. Sesuatu yang bantu lo memahami masalah, ngeliat fakta, nyari alternatif, ngitung risiko, dan nyusun rekomendasi.

Bukan pengganti otak lo. Tapi semacam sparring partner.

Konteks Apa Aja yang Harus Dikasih ke AI?

Nah ini bagian pentingnya.

Kalau lo cuma nulis “Peran: Manager Keuangan” di instructions, hasilnya tetep bakal generik. Karena “Manager Keuangan” itu cuma label. Bukan konteks.

Yang perlu dikasih ke AI itu kira-kira kayak gini:

• Jabatan — lo ini posisinya apa
• Unit kerja — lo di departemen apa
• Tujuan jabatan — kenapa posisi ini ada
• Tanggung jawab utama — lo bertanggung jawab atas apa aja
• Wewenang — lo boleh ngapain aja
• Batas kewenangan — sampe mana lo boleh putusin sendiri
• Atasan langsung — lo lapor ke siapa
• Bawahan — siapa aja yang di bawah lo
• Pihak yang sering diajak koordinasi — lo biasanya kerja bareng siapa
• Keputusan yang bisa diambil sendiri — mana yang gak perlu nanya
• Keputusan yang butuh persetujuan — mana yang harus lewat atasan
• Risiko utama — hal-hal apa yang paling rawan
• KPI — lo dinilai dari apa
• SOP atau kebijakan yang jadi acuan — aturan mainnya apa
• Gaya komunikasi — lo mau AI-nya ngomong kayak gimana
• Jenis bantuan yang diharapkan — lo pengen AI bantu di bagian apa

Keliatan panjang ya? Iya. Tapi ini yang bikin beda antara AI yang jawabannya generik sama AI yang jawabannya beneran kepake.

Gimana Sih Bentuk Konfigurasinya?

Kalau lo bikin custom GPT, di sana ada beberapa bagian yang bisa diisi.

Name. Ini nama GPT-nya. Misalnya “AI Assistant — Manager Keuangan”. Simpel.

Description. Ini penjelasan singkat. Misalnya: “Membantu Manager Keuangan menganalisis masalah, mempertimbangkan alternatif, memahami risiko, dan menyusun rekomendasi.”

Conversation Starters. Ini contoh pertanyaan yang bisa langsung diklik. Arahkan ke kasus nyata. Misalnya soal keterlambatan pembayaran vendor, disposisi Direktur, pengendalian biaya, permintaan pengeluaran besar, atau penyusunan bahan untuk Direksi.

Instructions. Nah ini yang paling penting. Di sini lo jelasin identitas lo sebagai Manager Keuangan. Posisi AI sebagai asisten berpikir. Konteks organisasi. Batas kewenangan. Cara analisis. Cara nanganin informasi yang belum lengkap. Mekanisme eskalasi ke atasan. Gaya bahasa. Format jawaban.

Dan yang paling krusial, di sini lo harus eksplisit bilang ke AI: “Jangan pernah menyuruh saya berkoordinasi dengan Manager Keuangan, karena saya sendiri Manager Keuangan.”

Kedengeran sepele. Tapi ini yang bikin jawaban AI gak jadi absurd.

Format Jawaban yang Enak Dibaca Itu Kayak Gimana?

Dari yang gw pelajarin, format yang paling masuk akal itu kayak gini:

• Inti Masalah — sebenernya masalahnya apa
• Yang Perlu Diperiksa — data atau fakta apa yang perlu dicek dulu
• Pertimbangan — hal-hal yang perlu dipertimbangin
• Alternatif — pilihan-pilihan yang ada
• Risiko — masing-masing pilihan risikonya apa
• Rekomendasi — saran yang paling masuk akal
• Langkah Berikutnya — habis ini ngapain

Kenapa formatnya kayak gitu?

Karena keputusan kerja itu gak sesederhana “iya atau tidak”. Ada konteks. Ada risiko. Ada wewenang. Ada orang lain yang terdampak.

Kalau AI cuma jawab “sebaiknya begini”, lo gak dapet apa-apa. Tapi kalau AI nunjukin jalan berpikirnya, lo bisa ambil keputusan dengan lebih sadar.

Bahasa AI-nya Harus Kayak Gimana?

Ini yang sering salah.

Banyak orang bikin AI assistant terus hasilnya bunyinya kayak robot korporat. “Berdasarkan analisis komprehensif, dalam rangka mengoptimalkan efektivitas sinergi lintas departemen...”

Buset. Bacanya aja capek.

Yang gw pengenin itu bahasa yang formal tapi natural. Profesional tapi gak kaku. Kayak ngomong sama rekan kerja yang pinter. Atau konsultan yang ngerti situasi lo dan ngomongnya gak bertele-tele.

Kalau di instructions, gw biasanya tulis sesuatu kayak: “Gunakan bahasa Indonesia yang profesional namun natural. Hindari frasa generik seperti ‘berdasarkan analisis komprehensif’ atau ‘dalam rangka mengoptimalkan’. Tulis seperti rekan kerja senior yang sedang membantu berpikir, bukan seperti laporan tahunan.”

Hasilnya lumayan beda.

Soal Knowledge, Boleh Gak Kasih Dokumen Perusahaan?

Hmm. Ini agak sensitif.

Kalau GPT-nya mau dipublikasikan, jangan. Jangan pernah kasih dokumen rahasia perusahaan ke GPT publik. Itu bahaya.

Versi publik sebaiknya dibuat generik. Framework-nya aja. Strukturnya. Profil jabatannya.

Kalau lo mau pakai versi personal buat internal, baru lo bisa masukin dokumen yang relevan. Tapi tetep sesuai kebijakan perusahaan lo. SOP, struktur organisasi, job description, delegasi kewenangan, pedoman keuangan. Itu bisa jadi referensi kalau memang boleh dan aman.

Intinya, jangan gegabah.

Fitur Apa Aja yang Perlu Diaktifin?

Dari yang gw coba, beberapa fitur ini berguna:

Web Search. Berguna buat hal-hal yang berubah-ubah kayak regulasi atau perpajakan. Karena aturan pajak kan gak bisa diam dari tahun ke tahun.

Code Interpreter atau Data Analysis. Ini potensinya bagus buat analisis Excel, cash flow, anggaran, laporan keuangan. Jadi lo bisa upload data terus minta AI bantu baca.

Image Generation. Buat versi awal, gak perlu.

Actions. Juga belum perlu di tahap awal.

Jangan semua fitur diaktifin cuma karena keliatan keren. Aktifin yang beneran kepake.

Sebelum Dipublikasiin, Harus Diuji Dulu

Ini yang sering dilupain orang.

Bikin GPT terus langsung publish. Gak dites. Ya jangan kaget kalau hasilnya aneh.

Gw biasanya bikin beberapa test case dulu. Misalnya:

• Kasus keterlambatan pembayaran vendor
• Target pengurangan biaya
• Permintaan pembelian aset bernilai besar
• Perbedaan pendapat dengan Direktur
• Permintaan penyusunan bahan untuk Direksi

Terus gw liat. Apakah AI-nya paham posisi gw? Apakah dia ngarang kewenangan? Apakah dia ngambil keputusan final sendiri? Apakah bahasanya masih natural?

Kalau ada yang gagal, ya revisi instructions.

Ini kayak ngetes karyawan baru sih sebenernya. Lo gak bisa langsung percaya dia handle proyek besar. Lo kasih tugas kecil dulu. Liat hasilnya. Baru naikin level.

Ini Bisa Jadi Produk Gak Ya?

Nah ini yang gw pikirin belakangan.

Awalnya gw cuma pengen bikin konten edukasi soal prompt engineering buat kalangan manajemen. Tapi lama-lama kepikiran, kok kayaknya ini bisa jadi produk ya?

Bukan cuma tutorial. Tapi semacam ekosistem. Misalnya gw kasih nama “JABAT.AI — AI Assistant untuk Setiap Jabatan”. Ada profil untuk Manager Keuangan, Manager HR, Manager IT, Manager Operasional, Direktur.

Tapi sebelum bikin banyak, mending bikin master framework dulu. Satu kerangka yang konsisten. Isinya Identity, Position, Objective, Authority, Responsibility, Organizational Context, Decision Boundary, Escalation, Risk Analysis, Communication Style, Decision Support.

Dari satu kerangka itu, lo bisa bikin profil untuk jabatan lain. Tinggal ganti isinya. Gak perlu mulai dari nol tiap kali.

Dan ini juga enak buat konten. Lo bisa bikin video yang nunjukin perbedaan antara AI yang ditanya tanpa konteks vs AI yang dikasih konteks jabatan dan kewenangan. Bedanya keliatan banget.

Pesan edukasinya sederhana: AI bukan sekadar perlu diberi pertanyaan. AI perlu diberi konteks tentang siapa kita, apa tujuan kita, apa kewenangan kita, dan di lingkungan organisasi seperti apa kita bekerja.

Satu Hal yang Perlu Diinget

Produk kayak gini sebaiknya jangan bergantung sepenuhnya sama satu platform.

GPT Store itu fitur yang bisa berubah. Mekanismenya bisa berubah. Ketersediaannya bisa berubah. Kalau lo bangun semuanya di atas satu platform, terus platformnya berubah, lo bisa kelabakan.

Makanya aset utamanya harus dibangun dulu. Framework. Profil jabatan. Instructions. Template. Knowledge structure. Test cases. Itu semua milik lo, gak bisa diambil platform.

GPT cuma salah satu wadah implementasinya. Bukan satu-satunya.

Jadi Intinya Apa?

Intinya gini. AI itu alat. Kayak pisau. Kalau lo cuma kasih pisau ke orang tanpa jelasin mau dipake buat apa, ya dia bingung.

Tapi kalau lo jelasin, “ini pisau buat motong sayur, kamu koki di restoran padang, sayurnya harus dipotong tipis-tipis”, baru dia bisa kerja bener.

AI juga sama.

Dia gak butuh pertanyaan yang pinter. Dia butuh konteks yang jelas.

Siapa lo. Jabatan lo apa. Wewenang lo sampe mana. Lo kerja di mana. Dan apa yang lo butuhin.

Kalau itu semua udah dikasih, jawabannya beda banget. Bukan lagi kayak artikel internet. Tapi kayak temen ngobrol yang ngerti situasi lo.

Entahlah. Mungkin kedengeran ribet di awal. Tapi sekali lo cobain, lo bakal ngerasa bedanya.

Lo sendiri udah pernah coba bikin custom GPT buat kerjaan? Atau masih pakai yang generik aja?

Pertanyaan yang Sering Muncul

Kenapa ChatGPT jawabannya selalu generik?

Karena lo cuma kasih pertanyaan, bukan konteks. AI gak tahu lo siapa, jabatannya apa, wewenangnya sampe mana. Jadi ya dia jawab pakai template umum.

Apakah harus bayar buat bikin custom GPT?

Buat bikin dan pakai sendiri, biasanya cukup pakai akun berbayar. Tapi detailnya bisa berubah tergantung platform, jadi cek langsung aja.

Boleh gak kasih dokumen perusahaan ke GPT?

Kalau GPT-nya publik, jangan. Bahaya. Versi publik sebaiknya generik. Kalau mau pakai dokumen internal, pastikan itu aman dan sesuai kebijakan perusahaan.

Berapa lama bikin GPT berbasis jabatan?

Kalau framework-nya udah ada, satu profil bisa kelar dalam hitungan jam. Yang lama itu mikirin isinya, bukan teknisnya.

Apakah AI ini bisa gantiin keputusan manajer?

Gak. Ini decision support, bukan decision maker. Keputusan tetap di tangan lo. AI cuma bantu lo mikir lebih terstruktur.

Building Job-Based AI Assistants: Why AI Needs to Know Who You Are (English Version)

Here's the thing. If you want to build a GPT or AI assistant that actually gives useful answers for work, the key isn't a clever prompt. The key is context. The AI needs to know who you are, what your position is, how far your authority goes, and what kind of organization you work in.

But wait. Before I go deeper into the framework, let me tell you why I even started thinking about this.

I had an experience that was both funny and annoying.

So I was trying to use ChatGPT to help me think through a problem. Let's say I'm a Finance Manager. There's an issue with vendor payments. Invoices are piling up, but cash flow is tight. I asked the AI something like: "I have a problem, vendor payments keep getting delayed, what should I do?"

And what did the AI say?

"You should coordinate with the Finance Manager."

Wow.

I AM the Finance Manager, buddy. Why are you telling me to coordinate with myself?

That's when it hit me. The problem isn't that the AI is stupid. The problem is the AI doesn't know who I am. It doesn't know my position. It doesn't know my authority. It doesn't know who my boss is, who my subordinates are, or what decisions I can make on my own.

All it knows is that I asked about vendors.

Of course the answer is generic. It's a Google result. A template answer. Something that sounds like a wikiHow article.

Why Does ChatGPT Always Give Generic Answers?

This question comes up a lot. And the answer is actually simple.

AI can only work with what you give it. If you only give it a question, it answers the question. But if you give it full context about your position, it can respond like a thinking partner who understands your situation.

Imagine you have a new assistant at the office. On their first day, you immediately ask: "What do you think we should do about the vendor problem?"

Can they answer well? No. They don't know who you are, what department you're in, what industry the company is in, or what the rules are.

But if you introduce yourself first. "Hey, I'm the Finance Manager at a distribution company, I have 3 staff, my approval authority goes up to 50 million, anything above that goes to the Director, and right now cash flow is tight."

Then they can give a relevant answer.

AI is the same.

What Is a Job-Based AI Assistant?

So here's the idea. Instead of using a generic AI that knows nothing about you, you build an AI assistant that's been given a "complete job description" of your position.

For example:

• AI Assistant — Finance Manager
• AI Assistant — HR Manager
• AI Assistant — IT Manager
• AI Assistant — Operations Manager
• AI Assistant — Director

And so on.

The GPT isn't meant to make final decisions. It's not meant to replace you. It's meant to be a thinking partner. Decision support. Something that helps you understand the problem, see the facts, explore alternatives, weigh risks, and draft recommendations.

Not a replacement for your brain. More like a sparring partner.

What Context Should You Give the AI?

This is the important part.

If you just write "Role: Finance Manager" in the instructions, the output will still be generic. Because "Finance Manager" is just a label. Not context.

What you need to give the AI is something like this:

• Position — what your role is
• Work unit — what department you're in
• Job objective — why this position exists
• Main responsibilities — what you're accountable for
• Authority — what you're allowed to do
• Authority boundaries — how far you can decide on your own
• Direct supervisor — who you report to
• Subordinates — who reports to you
• Regular counterparts — who you usually coordinate with
• Decisions you can make alone — what doesn't need approval
• Decisions that need approval — what has to go up the chain
• Main risks — what's most vulnerable
• KPIs — what you're evaluated on
• SOPs or policies you follow — what the rules are
• Communication style — how you want the AI to talk
• Type of help you expect — where you want the AI to assist

Looks long, right? Yes. But this is what makes the difference between generic AI answers and answers that are actually useful.

What Does the Configuration Look Like?

If you build a custom GPT, there are several fields you can fill in.

Name. The name of the GPT. Something like "AI Assistant — Finance Manager." Simple.

Description. A short explanation. Something like: "Helps Finance Managers analyze problems, consider alternatives, understand risks, and draft recommendations."

Conversation Starters. Sample questions users can click. Point them at real cases. Like vendor payment delays, Director's instructions, cost control, large expenditure requests, or preparing materials for the Board.

Instructions. This is the most important part. Here you explain your identity as a Finance Manager. The AI's position as a thinking assistant. Organizational context. Authority boundaries. How to analyze. How to handle incomplete information. Escalation mechanisms to your supervisor. Communication style. Response format.

And most crucially, you have to explicitly tell the AI: "Never tell me to coordinate with the Finance Manager, because I AM the Finance Manager."

Sounds trivial. But this is what keeps the AI's answers from becoming absurd.

What Response Format Works Best?

From what I've learned, the most sensible format is something like this:

• Core Issue — what the problem actually is
• What to Check — what data or facts need verifying first
• Considerations — what needs to be weighed
• Alternatives — what options exist
• Risks — what each option risks
• Recommendation — the most sensible suggestion
• Next Steps — what to do after this

Why this format?

Because work decisions aren't as simple as "yes or no." There's context. There's risk. There's authority. There are people affected.

If the AI just says "you should do this," you get nothing. But if the AI shows its reasoning, you can make a more conscious decision.

How Should the AI Talk?

This is where people often get it wrong.

Many people build an AI assistant and the output sounds like a corporate robot. "Based on a comprehensive analysis, in order to optimize cross-departmental synergy effectiveness..."

Ugh. Just reading it is exhausting.

What I want is language that's formal but natural. Professional but not stiff. Like talking to a smart colleague. Or a consultant who understands your situation and doesn't beat around the bush.

In the instructions, I usually write something like: "Use professional but natural English. Avoid generic phrases like 'based on a comprehensive analysis' or 'in order to optimize.' Write like a senior colleague helping you think, not like an annual report."

The difference is noticeable.

What About Knowledge? Can I Give It Company Documents?

Hmm. This is a bit sensitive.

If the GPT is going to be public, don't. Never give confidential company documents to a public GPT. That's dangerous.

The public version should be generic. Just the framework. The structure. The job profiles.

If you want a personal version for internal use, then you can include relevant documents. But still follow your company's policies. SOPs, org charts, job descriptions, delegation of authority, financial guidelines. Those can be references if it's allowed and safe.

Bottom line: don't be reckless.

Which Features Should You Enable?

From what I've tried, a few features are useful:

Web Search. Useful for things that change, like regulations or tax rules. Because tax rules don't stay the same year after year.

Code Interpreter or Data Analysis. Good potential for analyzing Excel files, cash flow, budgets, financial reports. You can upload data and ask the AI to help read it.

Image Generation. For an early version, not needed.

Actions. Also not needed at the start.

Don't enable everything just because it looks cool. Enable what you'll actually use.

Before Publishing, Test It First

This is what people often forget.

Build a GPT and immediately publish. No testing. Don't be surprised if the output is weird.

I usually create several test cases first. For example:

• Vendor payment delay case
• Cost reduction target
• Large asset purchase request
• Disagreement with the Director
• Request to prepare materials for the Board

Then I check. Does the AI understand my position? Does it invent authority? Does it make final decisions on its own? Is the language still natural?

If something fails, revise the instructions.

It's kind of like testing a new employee. You can't immediately trust them with a big project. You give small tasks first. See the results. Then level up.

Could This Become a Product?

This is what I started thinking about later.

At first I just wanted to create educational content about prompt engineering for management. But over time I thought, this could be a product, right?

Not just a tutorial. More like an ecosystem. Maybe call it "JABAT.AI — AI Assistant for Every Position." Profiles for Finance Manager, HR Manager, IT Manager, Operations Manager, Director.

But before building many, it's better to build a master framework first. One consistent skeleton. Containing Identity, Position, Objective, Authority, Responsibility, Organizational Context, Decision Boundary, Escalation, Risk Analysis, Communication Style, Decision Support.

From that one framework, you can build profiles for other positions. Just swap the contents. No need to start from scratch every time.

And this is also great for content. You can make a video showing the difference between AI asked without context versus AI given job and authority context. The difference is striking.

The educational message is simple: AI doesn't just need a question. AI needs context about who we are, what our goals are, what our authority is, and what kind of organization we work in.

One Thing to Remember

A product like this shouldn't depend entirely on one platform.

The GPT Store is a feature that can change. The mechanics can change. Availability can change. If you build everything on one platform, and the platform changes, you could be in trouble.

That's why the main assets should be built first. Framework. Job profiles. Instructions. Templates. Knowledge structure. Test cases. Those belong to you, and no platform can take them away.

GPT is just one implementation container. Not the only one.

So What's the Point?

The point is this. AI is a tool. Like a knife. If you just hand a knife to someone without explaining what it's for, they'll be confused.

But if you explain, "this knife is for cutting vegetables, you're a chef at a Padang restaurant, the vegetables need to be sliced thin," then they can work properly.

AI is the same.

It doesn't need a clever question. It needs clear context.

Who you are. What your position is. How far your authority goes. Where you work. And what you need.

If all that is given, the answers are completely different. No longer like an internet article. More like a conversation with a friend who understands your situation.

I don't know. Maybe it sounds complicated at first. But once you try it, you'll feel the difference.

Have you ever tried building a custom GPT for work? Or are you still using the generic one?

Frequently Asked Questions

Why does ChatGPT always give generic answers?

Because you're only giving it a question, not context. The AI doesn't know who you are, what your position is, or how far your authority goes. So it answers with a general template.

Do I need to pay to build a custom GPT?

To build and use one yourself, you usually need a paid account. But details can change depending on the platform, so check directly.

Can I give company documents to a GPT?

If the GPT is public, don't. It's dangerous. The public version should be generic. If you want to use internal documents, make sure it's safe and follows company policy.

How long does it take to build a job-based GPT?

If the framework is ready, one profile can be done in a few hours. The slow part is thinking through the content, not the technical side.

Can this AI replace a manager's decisions?

No. This is decision support, not a decision maker. The decision stays in your hands. The AI just helps you think more structured.

Terima kasih sudah mampir! Jika kamu menikmati konten ini dan ingin menunjukkan dukunganmu, bagaimana kalau mentraktirku secangkir kopi? 😊 Ini adalah gestur kecil yang sangat membantu untuk menjaga semangatku agar terus membuat konten-konten keren. Tidak ada paksaan, tapi secangkir kopi darimu pasti akan membuat hariku jadi sedikit lebih cerah. ☕️

☕ Buy Me Coffee