Memuat
Memuat
Want to go deeper?
Join one of our courses or ask us directly on WhatsApp.
Picked from the software, topic, and level of “Agentic AI Without Coding Class #50: Polishing and Presenting the Final Project, and Next Steps”.
Other ways we can help with Agentic AI.
Learn with a mentor
Scheduled AI & Machine Learning courses with live guidance and a certificate.

MATLAB Agentic AI Course: Building Autonomous AI Agents with MATLAB & Simulink

Midjourney Course: Create Stunning AI Images from Text
★ 4.9

DALL-E Course: Creating Professional AI Images from Text
★ 4.9
Have it built for you
No time to learn Agentic AI? Our team can do the project for you.

MATLAB & Simulink Development — Agentic AI for Simulation Automation and Data Analysis

AI Image Creation with Stable Diffusion — Unlimited Custom Visual Production
★ 4.9

AI Image & Video Creation with Leonardo AI — All-in-One Visuals for Business
★ 4.9
Get notified about new Agentic AI classes
At most one email a week. Unfollow anytime with one click.
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Welcome to the final episode of Agentic AI Without Coding. Your final project has been chosen and designed. Today we take it all the way to the finish line. We polish the project with a go-live checklist, practice a five-minute presentation, learn to give and receive feedback, go over the certificate criteria, and then close out our ten-session journey together.
Let's start with the go-live checklist. Area one, testing. Your project passes the test scenarios, including complaints and prompt injection tricks, like the ones we put together in episode forty-eight. Still on testing: try it on a phone or a real device, not just in the workflow editor. Many problems only show up when messages actually arrive in WhatsApp or Telegram. Area two, security. No customer data is pasted into prompts or into public AI. Accounts and API keys are held only by the business owner. If you shared an account with a friend while learning, now is the time to change the passwords and the keys. Area three, escalation. The path to a human has been tested, and there's a named person responsible, not just the admin. Area four, costs. API spending limits are set where available, and you've checked this month's bills, as in Session 9. Area five, documentation. There's a one-page document and a team SOP saved somewhere easy to find. And finally, the backup plan for when the bot fails has been tested, not just written down. If all eight items are ticked, your project is ready to be called live. For each area, prepare one piece of evidence you can show. For testing, for example, a test results sheet showing nineteen out of twenty scenarios correct. For security, the Session 9 security checklist, already ticked off. For escalation, a screenshot of the alert actually arriving on the monitor's phone. For costs, a monthly cost estimate taken from real bills, not a guess. And for documentation, the project document and SOP in one Google Drive folder. You'll use this evidence in your presentation too. It doesn't have to be perfect. If one area isn't ready yet, write in your thirty-day plan when it will be sorted out, and who will do it.
A project that only its builder understands will stop when its builder takes a day off. That's why we need documentation. AI can help write the draft. The role block: a documentation writer for a team that isn't technical. The data block: the project card from the previous episode, the list of nodes in your workflow, and the team SOP. The task block: write one-page documentation, so a new person can run and monitor this project. The format block: six sections with fixed headings. And a very important limits block: don't include passwords or API keys, and write ask the owner when the information isn't available. That way the AI doesn't make up content to fill the gaps. For Sunset Coffee, this documentation means Laras can step in for Dimas at any time. For Dewi, a teammate can monitor the report agent while she's on leave. Here are the six sections. First, the goal and problem: why this project was built. Second, how it works: a short flow plus a screenshot of the canvas. Third, accounts and tools: whose account it is, with no passwords. Fourth, who does what: the approver, the monitor, the data keeper, and the backup, as in the thirty-day plan. Fifth, if it fails: the backup plan steps. Sixth, the change history: the date, what changed, and who approved it. This last section is a huge help when the bot suddenly starts behaving differently. Check the AI's draft line by line before sharing it. Save it in the same folder as the SOP, and share the link with everyone involved.
Now the presentation. Every participant has five minutes, in the same order. Part one, the problem, about one minute. Mention one real number, for example eight hours a week spent just answering price questions. Part two, the solution, about one minute. Show the short flow on the canvas and name the tools you used. Part three, the demo, about a minute and a half. This is the most important part: show that your project really works. Part four, the impact, about one minute. Show the dashboard from episode forty-seven, and say that the numbers come from your own real records. Part five, the thirty-day plan, about thirty seconds: which stage is running now, and who the approver and monitor are. Notice that the order matches the project card. If your card is complete, the presentation is just a matter of reading it out with pride. Practice your presentation at least twice with a timer on your phone. Usually the solution part runs too long and the impact part is too short. Cut the technical explanations, because your audience cares more about results. The first fifteen seconds decide whether you have the audience's attention. Take Laras's opening line: night chats used to go unanswered, now they get a reply in one minute. That one sentence holds the problem, the solution, and a number all at once. Dewi could open the same way: the Monday report used to take two hours, now the draft is ready before I get to the office. Mr. Hendra could do it too: the student grade recap used to eat up my Saturday mornings, now the draft recap is ready every Friday afternoon. The problem, the solution, and a number, all in one breath. A few demo tips. First, use sample data. Don't show real customer names, numbers, or chats in front of other participants. Second, show one complete flow, for example from an incoming message to the order logged in Sheets. One complete flow is more convincing than five half-finished features. Third, also show the agent escalating to a human. That's exactly the part that makes the audience trust that your project is safe. Fourth, have a backup screen recording ready. The internet connection or a third-party service could fail right in the middle of your demo. Fifth, close with dashboard numbers, not a list of features. People remember impact, not node names. In class, presentations are usually scheduled like this. Each participant gets an eight-minute slot: five minutes to present and three minutes of feedback. Laras opens with the Sunset Coffee WhatsApp agent. Dewi follows with the weekly sales report agent, then Nadia with her content pipeline for several clients. Mr. Hendra from Bimbel Cerdas brings a weekly student grade recap, using the same pattern as the sales report. And in the last row, your own project. Look at the last column: every presenter has two feedback givers assigned in advance. That way, no presentation ends without input. If you're learning on your own, ask a coworker or a fellow business owner to be your audience.
After each presentation, the other participants give feedback. Nice and cool feel good, but they don't help much. It's better to say what was clear, for example the escalation part that made sense right away. The bot isn't smart enough is also too general. It helps more to be specific: when asked about shipping for two packages, free delivery wasn't applied. Instead of saying you should use a different tool, ask a question they can try: have you tried a separate stock tab, so it's easy to update? And don't just say nothing. Name one thing you'd like to copy. That's valuable feedback for the presenter too. When you're the one receiving feedback, remember that criticism is about the project, not about you. Other participants see your project with fresh eyes, and they often spot things you missed. To keep it easy, use the one, one, one pattern. First, one thing that worked, stated specifically. Second, one question that helps the presenter rethink, for example: what happens if someone forgets to update the stock in Sheets? Third, one suggestion that's small and can be tried within thirty days. Save big suggestions that need three months for later. This pattern keeps every piece of feedback short, polite, and useful. So how do you receive feedback? Listen to the end, take notes, and say thank you without getting defensive. After the session, Laras pastes the feedback notes into Claude and asks for the three points most worth acting on. Claude groups them. The point about shipping was mentioned by three participants, so it's the top priority. A separate stock tab and an opening message that mentions the admin's reply hours round out the top three. Big suggestions, like switching platforms or adding payments, are saved for a later month. Remember, the decision is still yours. Pick the ones with the most impact, then add them to week four of your thirty-day plan.
Now the certificate criteria. There are only two, matching the class goal we agreed on back in Episode 1. First, at least one agent or automation that's live, and actually used in your work or business. Second, a written thirty-day implementation plan. The five-minute presentation is how you show both. Notice that the criterion isn't the most sophisticated project. A simple automation used every day fully qualifies. So don't put off presenting just because your project feels too simple. What counts is whether the project is alive and has a plan to keep running. For the first criterion, prepare a demo recording and a real execution history, such as the execution history page in n8n or Make, or the chat history of your AI assistant. This history shows the project is really in use, not just tried once. For the second criterion, prepare the plan document: stages and pass criteria, roles, the backup plan, and the SOP. You already put all of this together in episode forty-eight. How and when to submit follows the instructions from your class organizer. Keep all the evidence in one folder so it's easy to send. To keep things tidy, gather everything in one final project folder. First, the automation map with scores and the main project from episode forty-six. Second, the impact dashboard with before and after numbers from episode forty-seven. Third, the thirty-day plan, with roles, the backup plan, and the SOP from episode forty-eight. Fourth, the project card from episode forty-nine and the one-page documentation we put together today. Fifth, the demo recording and the real execution history. This folder isn't just for the certificate. It's a working asset you can open again whenever the automation needs fixing or growing.
Before we part ways, let's look back at our journey. In Session 1, we cleared up the difference between chatbots, automation, and AI agents, then mapped out your five repetitive tasks. In Session 2, we learned prompting that gives consistent results, and put together ten ready-to-use prompt templates. In Session 3, we built a custom AI assistant that knows your business's products, prices, and voice. In Session 4, the AI started searching the internet on its own and acting in the browser, with rules for when it must ask for approval. In Session 5, we connected AI to Gmail, Calendar, and Sheets through connectors, while keeping data access permissions in check. In Session 6, your first automated workflow ran in n8n or Make: a form comes in, the AI summarizes it, it's saved to Sheets, and an email goes out. In Session 7, we gave the agent tools and memory, until it could answer stock and price questions and log orders. In Session 8, that agent served customers twenty-four hours a day on WhatsApp and Telegram, with escalation to a human. In Session 9, we built an agent team of researcher, writer, and reviewer, with owner approval, safeguards against AI making things up, and a security checklist. And in Session 10, it all came together into a final project: the automation map, impact measurement, the thirty-day plan, and the presentation. From understanding to running, all without writing a single line of code.
So what comes after this class? First, carry your thirty-day plan through to the end. One project that's truly live is worth more than five ideas that never get off the ground. Second, remeasure every month. Update the impact dashboard and make a decision: keep, fix, or replace. Third, take the next automation from your automation map. Start with the quick wins, then break big projects into small stages. Laras, for example, has already planned the weekly sales recap for next month. Fourth, bring your team along. Share the assistants, documentation, and SOPs, so the automation doesn't depend on just one person. Fifth, keep up with the tools. Menu names, features, and prices will keep changing, so check the official pages regularly. But the principles you've learned still apply: start from a real problem, choose the simplest way, keep human approval, and measure the results. Let's count what you're taking home. Ten ready-to-use prompt templates from Session 2. One custom AI assistant that knows your business, from Session 3. At least one workflow or agent that's live and in use, from Sessions 6 to 10. And one thirty-day implementation plan that's ready to run. Plus an automation map that still holds the next tasks waiting to be automated. And here we are at the end of the class. Remember the goal we agreed on in the first episode: one agent or automation that you actually use. If your project is live, you've reached that goal. Congratulations. The class may be over, but your business automation journey is just beginning. Every hour the agent saves is an hour you can spend on your customers, your family, and new ideas.
Let's recap this ten-session journey. You learned how agents work, wrote consistent prompts, built an assistant, connected AI to your work apps, built workflows and agents, served customers on WhatsApp, protected quality and security, then put it all into practice with a final project and a thirty-day plan. Thank you for learning with Cerdas Inovasi Digital in Agentic AI Without Coding. Congratulations on your final project. May the agents and automations you've built truly lighten your work every day. Wishing you every success in your business and your career.
Try these on your own after watching to lock in what you learned.
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
Notes are saved with the current video time and are only visible to you.
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.

Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
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.
See our clients →