PTENES
MODULE 1.2

⚖️ No-code automation vs. AI-powered apps

Two ways to solve problems with AI: visual workflows with fixed steps and agentic development that writes code. When to use each—and why mastering both makes you unbeatable.

6
Topics
35
Minutes
Basic
Level
Theory
Type
1

🔁 Deterministic Automation (Make/n8n)

A deterministic automation is a flow of fixed steps: you visually design a sequence that always happens in the same order— trigger → action → action. Tools like Make and n8n are built exactly for this: drag blocks and connect them.

🧭 Main concept

"Deterministic" means no surprises: the same trigger always runs the same steps. You don’t ask the tool to “think” — you lay out the entire path in advance.

  • •Trigger — what gets it started (a submitted form, a time, a new row in a spreadsheet).
  • •Modules — each block does one thing (send an email, save a record, call an API).
  • •Webhook — an address that receives data from outside and starts the flow right away.
Trigger form submitted Module 1 transforms data Module 2 save a record Final action sends email

💡 Practical tip

If you can draw the entire task on one sheet as a sequence of arrows that never changes, it's a perfect candidate for no-code. Cheap, fast, and easy to maintain.

Trigger

the starting point

Modules

action blocks

Webhook

external data

Fixed

always the same

2

🤖 Agentic development (AntiGravity)

In agentic development, you don't draw the path: you describe the goal. The AI then writes the code, runs it, sees the error that appeared, and fixes it on its own — repeating the cycle until it works. It’s reasoning, not a script.

1

You describe the goal

"I want an app that accepts a spreadsheet and generates a summary report." Don't say how—just what.

2

AI writes and tests

It generates the code, runs the project, and observes the actual result—as a developer would.

3

Sees the error and fixes it

If something breaks, it reads the error message, adjusts the code, and tries again. This loop is what helps solve complex problems.

That’s why the feature is practically unlimited: since the output is real code, you don’t run into the limits of a closed set of blocks. If it can be programmed, it can be built.

Goal

you describe

Writes

the AI codes

Fixes

test and adjust

Unlimited

no fixed blocks

3

📦 "Everything Is JSON Under the Hood"

Here’s a secret that demystifies everything: that pretty screen for dragging blocks is just a interface over code. Under the hood, the entire workflow is described in JSON — a structured text format that lists each block and each connection.

the same workflow, in JSON
{
  "gatilho": "formulario_enviado",
  "passos": [
    { "acao": "transformar_dado" },
    { "acao": "salvar_registro" },
    { "acao": "enviar_email" }
  ]
}

In practice, you can copy a workflow from a visual tool and paste it—and what gets passed along is precisely that JSON. In other words: you were already working with “code” without realizing it. That takes away the fear of going down a layer and working closer to the foundation.

🔍 Try it

In a no-code tool, copy a scenario and paste it into a text editor. You’ll see the structure in JSON. You don’t need to understand every line — just notice that "code" is simply organized text.

Interface

the nice-looking screen

JSON

the text behind it

Copy

you see the structure

Demystifies

less fear

4

🎯 Which Tool for Which Problem

There is no abstractly "best" tool — there is the right for the problem. Two questions resolve almost every doubt: "does this change?" e "does it need its own screen?".

✓ Use no-code when

  • ✓The task is simple and stable
  • ✓The steps are always the same
  • ✓Doesn’t need its own interface
  • ✓Speed is more important than customization

✗ Prefer an agentic app when

  • ✗The problem is complex and changes
  • ✗The logic is custom and nontrivial
  • ✗You need a custom screen
  • ✗Full control matters more than speed
QuestionIf NOIf YES
Does this change over time?→ no-code→ agentic app
Does it need its own screen?→ no-code→ agentic app
Is the logic complex?→ no-code→ agentic app
Does it change?

the 1st question

Screen?

the 2nd question

Stable

→ no-code

Own

→ app

5

🧩 Why learn both

No-code and agentic building don’t compete — they complete. If you understand how a step-by-step workflow works, you can guide an agent better (you know what to ask and in what order). And if you've already built an app, you have a better sense of when a simple workflow is enough.

✓ Knowing no-code helps with agentic

  • ✓You think in clear steps
  • ✓Describe the goal more precisely
  • ✓Recognizes triggers and actions

✓ Knowing agentic helps with no-code

  • ✓You see the limits of the fixed workflow
  • ✓Knows when to "raise the bar"
  • ✓Doesn’t force an app where one isn’t needed

🔌 They also connect

The two worlds communicate via MCP (a standard that connects tools and agents). An app you built can be triggered by a no-code workflow—and vice versa. A broader toolkit means more solutions at your disposal.

Complementary

don’t compete

Drive

thinks in stages

Expertise

more solutions

MCP

integration

6

🏠 IKEA vs. Building from Scratch

An analogy brings the idea together: the cloud vibe coders (platforms that build and host for you) are like a IKEA house. They're ready to use and quick to set up—but they have fixed structures. Building in local environment is building the house however you want, wall by wall.

🏠 IKEA Home (Cloud)

  • +Ready to go and fast — starts in minutes
  • +No need to manage the foundation
  • −Fixed structures — you can only change so much
  • −You depend on the platform

🧱 Build from scratch (local)

  • +Full control — any plant is possible
  • +The code is yours and stays on your machine
  • −Requires more from the builder (you + the AI)
  • −Less "ready-made"

⚖️ The trade-off in one sentence

Convenience on one side, control on the other. There’s no one right answer: start with IKEA when you want speed; go local when you hit the limits of fixed structures.

IKEA

ready-to-use cloud

From scratch

free space

Convenience

speed

Control

total freedom

🔬 Practical example

The same idea — "notify me when a price drops" — it can live in different layers, depending on what you want to deliver:

  • 1.Just notify? "Checked the price → sent a message." Fixed steps → no-code solve.
  • 2.Want a dashboard with login, history, and charts? This is its own screen with its own logic → local app (agentic).

Notice: the idea doesn’t determine the tool — the level of complexity and whether you need a custom interface do.

🏋️ Exercise

Get a automation you would build in Make (something simple and fixed) and describe it as a agentic app would do differently. For each version, answer:

  1. What the app wins compared with the no-code workflow (screen, history, flexibility)?
  2. What it costs more (build time, maintenance, complexity)?
  3. In the end: is it worth moving up a layer, or is the simple workflow enough?

This exercise trains you to choose the right tool before you start building.

⚡ Ready-made prompts

Copy, paste into your AI assistant, and adapt to your use case.

compare the two approaches
Compare como "[tarefa]" seria feita como fluxo no-code (Make/n8n)
vs. como app agentic (um IDE com agentes de IA).
Liste prós e contras de cada abordagem em tópicos curtos
e termine recomendando qual usar e por quê.
explain with an analogy
Sou iniciante. Explique, com uma analogia simples, a diferença entre
automação determinística (passos fixos, tipo Make/n8n) e construção agentic
(a IA escreve e corrige o código). Use o contexto de "[meu negócio]".

✅ Module summary

✓
Deterministic — visual flow with fixed steps (trigger → action); cheap and simple for predictable tasks.
✓
Agentic — you describe the goal, and the AI writes, tests, and fixes it; unlimited functionality.
✓
It’s all JSON under the hood — the visual screen is just an interface over code; that demystifies “programming.”
✓
Choose based on the question — “does this change? Does it need a screen?”; learning both expands your skill set.

Next module:

1.3 — Complete environment setup