At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Welcome to Agentic AI Without Coding, a class from Cerdas Inovasi Digital. In this first episode we clear up three terms that often get mixed up: chatbots, automation, and AI agents. After watching, you'll be able to judge for yourself which tasks fit each one.
Step 1: What we will build in this class
Before we start, let's agree on the goal. This class is not a tour of AI app features. The goal is that by the end of the class, you have at least one agent or automation that you actually use in your own work or business, and that saves you real time every week. This class is designed for five groups. First, small business owners who reply to customer chats, log orders, and create sales content every day. Second, office professionals who are drowning in emails, meeting schedules, and routine reports. Third, marketers who need to research the market, write captions, and plan content calendars quickly. Fourth, admins who process form data, put together recaps, and send reminders. And fifth, teachers who prepare lessons, quizzes, and feedback for many students at once. If your work looks like one of these, you're in the right class. The requirements are simple. First, a laptop or PC with a modern browser such as Chrome, Edge, or Firefox. Second, a Google account. We'll use Gmail, Google Drive, Sheets, Calendar, and Forms a lot. Third, a ChatGPT or Claude account. The free version is enough for the early sessions, and I'll tell you when a paid feature is really needed. Fourth, an n8n or Make account for visual automation. You'll only need it from Session 6, and both have free plans for learning. One more important thing: you won't need to write a single line of code. Our journey has four stages. Sessions one and two build understanding: what an agent is, and how to give instructions that produce consistent results. In sessions three to five, we create an AI assistant for your business, let AI research and take action in the browser, and then connect it to Gmail, Calendar, and Sheets. Sessions six to eight are the heart of automation: visual workflows in n8n or Make, agents that have tools, and customer service on WhatsApp and Telegram. Finally, sessions nine and ten cover quality, security, and costs, followed by your own final project. Every session produces one real output that you reuse in the next one, so keep all your practice results.
Step 2: Chatbots: AI that answers when asked
Let's start with the most familiar one: the chatbot. When you open ChatGPT, Claude, or Gemini and type a question, you're using a chatbot. For example, a coffee shop owner asks for three Instagram captions for a palm sugar iced coffee promo, priced at eighteen thousand rupiah, valid Friday to Sunday. Within seconds, the chatbot writes three neat caption options, complete with the price and the promo period. For writing tasks like this, chatbots are very strong. Now look at the next request. The shop owner asks the chatbot to post the second caption straight to the shop's Instagram. The answer: it can't. A chatbot only writes inside the chat window. Copying the text, opening Instagram, and pressing the post button is still up to you. A chatbot always works the same way. First, you ask a question or give an instruction. Second, the AI answers. The answer can be text, a table, a list of ideas, or a summary. Third, the AI stops and waits. It won't do anything else until you write your next message. Fourth, you take the action yourself: copying the result, sending the email, or entering the data. So the key is this: you start it, and you take the action. AI helps with thinking and writing in between. To be fair, let's look at the strengths and weaknesses. When it comes to writing, chatbots are amazingly fast. The weakness is that the writing style can vary when instructions are unclear. We fix that in Session 2, on prompting. When it comes to thinking, chatbots are good at summarizing, explaining, and giving options. But they can sound very sure while being wrong. This is called hallucination, or making things up. We cover how to prevent it in Session 9. When it comes to acting, a regular chatbot can't do anything. It doesn't open other apps, and it doesn't send anything. And as for memory, a chatbot remembers what's in one conversation, but in a new conversation it forgets who you are and what your business does. We solve this problem in Session 3 with a custom AI assistant.
Step 3: Automation: if-then rules that run on their own
The second way is automation. Imagine an online shop that takes orders through a form. For every order, the admin has to copy the data into a spreadsheet and then let the owner know. With automation, we create a rule. If a form comes in, which we call the trigger, then add a new row in Google Sheets. After that, send the admin an email with the order details. Then send a WhatsApp notification to the owner. These four steps run on their own, day or night, with nobody watching. This is what a workflow looks like in a visual automation app such as n8n or Make. Each box is called a node or a module, and the arrows show the order in which data flows. We'll build a workflow like this ourselves in Session 6. You can see the result in Google Sheets. Every form that comes in instantly becomes a new row: time, name, product, quantity, and notes. Consistent, and it never forgets. But look at the note on Sari's order: please send it tomorrow, I'll pay by transfer. The automation still treats it like a normal order. And Andi's order actually contains a question: is there a sugar-free version? The automation doesn't understand that this is a question that needs an answer. It just copies the text into the notes column. So the strength of automation is consistency. The steps are exactly the same every time, and it never forgets. The weakness is that it's rigid and only follows the rules we set. Automation runs on its own, twenty-four hours a day, once an event triggers it. But it doesn't understand unexpected messages, like a request to deliver tomorrow or a customer's question. Automation is also good at connecting apps: Forms, Sheets, Gmail, WhatsApp, and hundreds of others. The catch is that every new possibility has to be added as a rule by hand. In short, automation is a pair of hardworking hands, while a chatbot is a brain that writes well. So what happens when we combine the two?
Step 4: AI agents: given a goal, they decide the steps and use tools
This is the third way, and the heart of this class: the AI agent. The simple formula is this: an agent is an AI brain, plus tools, plus a goal. The difference shows in how you give instructions. With automation, you set up every step. With an agent, you just give it a goal, and the agent works out the steps on its own. Let's look at an example with the same order data. The shop owner gives the agent a goal, not steps: check this morning's orders, make sure they can all be processed, and prepare replies for customers who asked questions. The agent gets to work. First, it uses a tool to read the orders spreadsheet and finds four new rows. Then, without being told, it opens the stock spreadsheet, because to make sure the orders can be processed, it needs to know what's in stock. This is what we mean by making its own plan. The result isn't just a copy of the data. The agent understands that Sari wants delivery tomorrow, and that Andi is asking a question. It even finds a problem we didn't ask about: only four bottles of iced coffee are left, but today's orders need five. Notice also that before changing the delivery date, the agent asks first: approve? A good agent asks for permission before taking an important action. This principle is called human in the loop. The shop owner approves. The agent then uses its next tool to change the delivery date in the spreadsheet. Then it sends a WhatsApp reply to Andi. Finally, the agent reports back and offers a sensible next step: setting a reminder to make more coffee. From a single goal, the agent carried out six steps while still asking for permission at the important point. Behind the scenes, an agent like this is built from a few parts. The first part is the trigger: for example, a message coming in from a chat or WhatsApp. The second part is the AI Agent node itself, which holds the goal and your work instructions. The third part is the AI model, which acts as the brain, for example a model from ChatGPT or Claude. The fourth part is tools. A Sheets tool to read and update data. A WhatsApp tool to send replies. And memory, so the agent remembers what was said earlier in the conversation. We'll build exactly this setup in Session 7, without writing any code.
Step 5: Comparing chatbots, automation, and agents
Now let's line up all three. First row: who starts it. A chatbot is started by you. Automation is started by an event, such as a form coming in. An agent is started by a goal from you, or by an event too. Second row: who decides the steps. With a chatbot, you do. With automation, it's the fixed rules set up at the beginning. With an agent, the agent works out the steps itself. Third row: can it act in other apps? A regular chatbot can't. Automation can, but only following its rules. An agent can, and it chooses for itself which tools to use. Fourth row: does it understand messages? Chatbots and agents understand human language. Pure automation doesn't. And the examples: a chatbot for writing captions, automation for moving forms into a spreadsheet and an email, and an agent for checking orders, replying to customers, and asking for permission at the important points. In practice, the three aren't separate boxes with sharp edges. They form a spectrum. On the far left is the chatbot, which only answers. One level up is automation with an AI step. The flow is fixed, but in the middle, AI summarizes or categorizes the message. We build this in Session 6, and it's already very useful. The next level is an agent with approval. The agent is allowed to act, but it asks for permission before sending, changing important data, or paying. On the far right is an independent agent that acts fully within the limits you set. The further right you go, the more time you save, but the limits and supervision must also get tighter. For a small business, the sweet spot is usually in the middle. To make this concrete, picture a shop that gets forty customer chats a day. Replying by hand, one by one, including checking stock and prices, can take about three hours. With a chatbot's help, the answers are written faster, but they still have to be copied and pasted by hand. That's maybe about an hour and a half. With an agent connected to the stock data and WhatsApp, most chats are answered automatically, and you only approve the special cases. The time can drop to about thirty minutes. These numbers are an illustration, and in Session 10 you'll measure the real savings in your own business.
Step 6: Choosing the right one for your work
So how do you choose? Use three deciding questions. The first question: are the steps always the same every time? If yes, like moving form data into a spreadsheet, start with automation. The second question: does the task need understanding, judging, or writing? For example, reading customer complaints or writing a summary. If yes, bring in AI, either through a chatbot or an AI step inside the automation. The third question: does it need to decide the next step on its own, and use several apps at once? If yes, it's an agent candidate, with approval points, of course. The key principle: choose the simplest option that solves the problem. Agents are exciting, but for moving data around, regular automation is cheaper, faster, and easier to check. Let's practice. The first task: writing ten content ideas for next month. The answer is a chatbot. It's a one-off job, it needs creativity, and you still pick the results yourself. Second: every sign-up form goes into a spreadsheet, and the participant gets a confirmation email. That's automation, because the steps are always the same. Third: summarizing customer emails and adding priority labels. That's automation with an AI step. The flow is fixed, but the content of each email has to be understood. Fourth: answering stock and price questions, then logging the order. That's an agent, because it has to decide on answers and use several tools. Exactly what we build in Session 7. Fifth: researching competitor prices every week and turning them into a report. That's also an agent, one that searches, compares, and writes. We build it in Session 4. Before moving on to the next episode, do this small task. First, write down three repetitive tasks from your business or office. Second, answer the three deciding questions for each of those tasks. Third, mark your choice: chatbot, automation, automation with an AI step, or agent. Fourth, keep your notes. In Episode 5 we turn them into a list of five tasks to automate, and that list will guide your project until the end of the class.
Done!
Let's recap. A chatbot answers when asked. Automation runs fixed rules when triggered. And an AI agent is given a goal, works out its own steps, uses tools, and then checks the results. In the next episode we look at how an agent thinks: from goal, to plan, to action, to checking results. Understanding this will make it much easier for you to design your own agents. See you in episode two.
Practice exercises
Try these on your own after watching to lock in what you learned.
Write down three tasks from your business or office, then mark which ones a chatbot can handle, which fit automation, and which need an agent.
For each task, answer the three deciding questions: are the steps always the same, does it need judging or writing, and does it need several apps?
Keep these notes. In Episode 5 we turn them into a list of 5 tasks to automate.
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
52 video classes1,276 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
