Your First AI Automation

Last updated: Jan 2026

Overview

You've learned the concepts. Now let's put them into practice. In this guide, we'll build a real-world AI automation from scratch using ORCFLO's visual builder, walking through the design process, implementation, and iteration.

By the end, you'll have a working automation and the skills to create your own custom AI workflows.

What We'll Build

We'll create a Content Repurposing Workflow that takes a blog post and automatically generates multiple content pieces for different platforms.

Content Repurposing Workflow

  1. Input Node: Blog Post - Define input fields for the original article
  2. LLM Step: AI Processing - Claude analyzes and creates variations
  3. Output Node: Multiple Formats - Twitter thread, LinkedIn post, email newsletter in markdown

Why This Example?

Content repurposing is a common task that showcases AI strengths: understanding context, adapting tone, and generating creative variations.

Design Phase

Before building, let's design our workflow. Good design saves time and prevents rework.

1. Define Input Fields

What data does the workflow need to start? Configure your Input node:

  • blog_content - The article to repurpose (required)
  • author_name - For attribution (optional)
  • brand_voice - Tone preferences (optional)

2. Define Outputs

What should the workflow produce? Output node formats results as markdown:

  • Twitter thread - 5-7 connected tweets
  • LinkedIn post - Professional format
  • Email excerpt - Newsletter-ready summary

3. Map the Process

How should data flow through the workflow?

text
Input Node → LLM Step → Output Node
     ↓
[blog_content, author, voice]
     ↓
[AI: Analyze & Generate]
     ↓
[Markdown: All formats combined]

Building the Workflow

Now let's build the workflow step by step using ORCFLO's visual canvas.

1

Add Input Node

Drag an Input node onto the canvas. Add a Long Text field named "Blog Post Content" (required) and a Text field named "Author Name" (optional).

2

Add LLM Step

This AI model node will generate all content variations. Select Claude Sonnet 4 as the model.

3

Add Output Node

Configure the Output node to collect the AI's response in markdown format. The output from connected LLM steps flows in automatically.

4

Connect Nodes

Drag connections from Input to LLM Step to Output. The data will flow through automatically.

Configuring the AI

The prompt is crucial to getting good results. Write clear instructions in plain text describing what you want the AI to do. The input data from connected nodes flows in automatically.

LLM Step Prompt
Given the blog post provided, create three content variations:

Please generate:

### 1. Twitter Thread (5-7 tweets)
- Start with a hook that grabs attention
- Each tweet should be under 280 characters
- Use relevant emojis sparingly
- End with a call-to-action

### 2. LinkedIn Post
- Professional tone
- Include a compelling opening line
- Use 2-3 paragraphs
- Add 3-5 relevant hashtags

### 3. Email Newsletter Excerpt
- Friendly, conversational tone
- 150-200 words
- Highlight the key takeaway
- Include a clear CTA

Format your response with clear headers for each section.

Prompt Best Practices

PracticeDescription
Be specific about formatTell the AI exactly how to structure the output
Set constraintsCharacter limits, word counts, and number of items
Define toneProfessional, casual, friendly - be explicit
Request structureUse headers and sections for parseable output

Testing and Iterating

Now let's test the workflow. Click the Run button.

1

Enter sample data

A dialog appears for input values. Paste a real blog post.

2

Watch visual feedback

Nodes turn green on success, red on failure in real-time.

3

Inspect with execution panel

Click nodes to see inputs, outputs, tokens used, time, and cost.

4

Refine the prompt

Adjust instructions based on what works and what doesn't.

Iteration is Normal

It's rare to get perfect results on the first try. Plan to iterate on your prompts 2-3 times to optimize the output quality.

Going Further

Now that you have a working automation, here are ways to enhance it using ORCFLO's additional node types:

EnhancementDescription
Add conditional branchingUse a Decision Point to verify quality meets standards before output
Multiple LLM StepsUse different AI models for each platform variation

Next Steps

  • Add scheduled executions for recurring content
  • Add Note Cards to document workflow logic
  • Try different AI models for different tasks

Key Takeaways

  • Design before building: map input fields, processing, and output format
  • Use Input nodes for data entry, LLM Steps for AI, Output for markdown results
  • Write detailed, structured prompts in plain text - data flows between nodes automatically
  • Test with Run button and watch real-time visual feedback
  • Iterate on prompts and add Decision Point for advanced logic
← Understanding Workflows