This comprehensive LinkedIn automation playbook is designed to streamline your content creation by taking viral news stories, conducting additional research, writing LinkedIn posts in your specific tone of voice and brand guidelines, and even generating AI images for you.
It provides you with complete control over everything that gets published.
Here’s how to get it up and running:
Step 1: Lay the Groundwork with Planning and AirTable Setup
Before diving into the build, it’s helpful to plan the high-level flow to identify potential errors or necessary logic.
1. Understand the Core Flow: At a high level, the system gets a news article, does additional research, writes the article (post), creates an image, posts it to LinkedIn, and then stores it back in your database for reference.
2. Detailed System Flow:
- The system will run on a schedule.
- It will get viral news articles and store them in your AirTable database.
- AI will research the articles.
- It will fetch your brand guidelines and tone of voice.
- It will generate the LinkedIn post copy.
- It checks a checkbox to see if an image is needed.
- If an image is needed, it generates an image.
- It then updates the database with either the post and image, or just the post content.
- Finally, it posts to LinkedIn and marks the post as complete.
3. Set Up AirTable as Your Central Hub: AirTable is the “hub for the entire system” and provides “total control over everything that gets published”. You’ll need to create an AirTable base with fields to store:
- Source headlines.
- Article summaries.
- Links to the original articles.
- Social channel selection (e.g., a “LinkedIn” checkbox/field).
- An “image needed” field (yes/no).
- The generated LinkedIn post copy.
- The generated AI image URL/attachment.
- Status fields (e.g., “waiting for content”, “approved”, “posted”).
- A separate table/section for your Brand Guidelines and Tone of Voice.
Step 2: Implement News Scraping and Initial AirTable Population
The system is built around viral or trending news sources, which are sourced by scraping RSS feeds.
1. Configure RSS Feed Scraping: Set up a component within your automation tool (like n8n’s RSS feed node) to scrape RSS feeds that you choose (e.g., about artificial intelligence).
2. Extract and Summarise Articles: This component will:
- Get all articles from the RSS feeds.
- Check for existing URLs to prevent adding duplicate content.
- Pull the HTML/text from the articles.
- Use an AI model to summarise each article automatically.
3. Store in AirTable: The scraped and summarised information, including the source headline, summary, and a link to the original article, is then stored back inside your AirTable database. This populates your AirTable with current news.
Step 3: Build the LinkedIn Post Generation Workflow
This is where the magic happens, transforming news into personalised LinkedIn content. This section runs after you’ve reviewed the scraped content in AirTable and made your selections.
1. Retrieve Selected Articles from AirTable: Your workflow should search AirTable for records that have the status “waiting for content”. Crucially, it must also check if the “LinkedIn” social channel field is selected.
2. Loop Through Articles: Implement a loop to process each selected article one at a time. This provides AI agents with “time to breathe” and ensures clarity for what they should be talking about, preventing them from getting “stuck between two different posts”.
3. Generate Research Queries (AI Agent): Input: The news article’s headline and summary from AirTable. AI Prompt: Use an AI agent with a prompt like: “You are a search query generator for a LinkedIn research system. Given a source headline and summary, output four targeted queries less than 15 words each to gather diverse high-quality data.”.
Output: Four targeted search queries designed to find: recent statistics, case studies, expert opinions, and competing content/common questions related to the article topic.
4. Conduct Additional Research (Research Agent): Input: The generated search queries. Tool: You can use an HTTP tool, ChatGPT, Perplexity, or another research tool. AI Prompt: A prompt like: “You’re a research agent in a multi-agent blog creation system. Your task is to gather concise high-quality information for a keyword using the provider research and output it in a condensed format,” with rules for bullet points, key insights, implications, and statistics.
Output: Concise, high-quality information, including citations, to make your LinkedIn post that much more compelling.
5. Fetch a Random Writing Framework: Hardcode (or ideally, store in AirTable) several writing frameworks (e.g., Problem Agitate Solve (PAS), Insight Impact Recommendation, Story Insight Action).
These are essentially prompts for the AI agent. Implement logic to randomly select one framework for each post. This provides “variety in your content”. Each framework defines the post’s purpose, guidelines, structure, and formatting.
6. Retrieve Brand Guidelines. Query AirTable to fetch your personal brand guidelines and tone of voice.
Ensure your brand guidelines are detailed enough but concise enough to prevent the LLM from getting confused. These can be easily updated in AirTable. Craft the LinkedIn Post (LinkedIn Writer AI Agent): This is the core writing component. System Prompt: Feed the randomly selected writing framework as the system prompt (e.g., “You’re an expert on LinkedIn and skilled at crafting impactful copy using the [Framework Name] framework…”).
User Prompt: Craft a user prompt that instructs the AI to “craft a LinkedIn post based on the following news story and additional research focus on the article headline and summary”.
Input Data: Pipe in the: Article headline, original summary, additional research from your research bot, your tone of voice guidelines (e.g., “Be casual, spartan, and use normal language. Please follow these tone of voice guidelines…”).
Output: A compelling LinkedIn post that combines the news story, additional research, writing framework, and your tone of voice. This combination ensures “a lot of variety” and prevents posts from becoming “boring or repetitive or redundant”.
8. Store Post in AirTable: The generated LinkedIn post copy is saved back into your AirTable database.
Step 4: Implement AI Image Generation (If Selected)
This step is conditional, only running if you’ve selected “yes” for an image in AirTable.
1. Conditional Check: Use an “if node” or similar logic to check the “needs image” field in AirTable. If “yes,” proceed with image generation; otherwise, skip to updating the database with just the text.
2. Identify Content Type (AI Agent): Input: The article summary. AI Prompt: Instruct an AI agent to “identify the main theme of a given article” choosing from predefined categories (e.g., conceptual, human, strategic, symbolic, reflective). Output: The identified content type (e.g., “strategic post”).
3. Select Visual Style: Based on the identified content type, retrieve a corresponding visual style guide. These can be hardcoded or stored in AirTable similar to writing frameworks.
Examples: “editorial flatlays or still lives” for strategic posts, “lifestyle photography” for human posts, “soft minimalistic illustration” for conceptual posts. This builds consistency and brand recognition.
4. Generate Image Prompt (AI Agent): Input: The article summary and the selected visual style guide. AI Prompt: “You’re an AI generated vivid image prompts for a specific article based on its summary. Do not output any special characters.” Include examples and constraints.
Output: A vivid image prompt tailored to the article’s theme and desired visual style (e.g., “editorial flat layer still life artfully arranged work desk showcasing AI interaction tools laptop with code snippets energy efficient light bulb”).
5. Generate Image (DALL-E 3 or similar): Feed the generated image prompt into an image generation service (like OpenAI’s DALL-E 3). Ensure the image generator is configured to respond with image URLs, not binary data files.
6. Wait Timer: Include a short wait timer (e.g., 5 seconds) after image generation. This allows the image to fully register on the server before attempting to store it.
7. Store Image in AirTable: The image URL is saved back into your AirTable database, ensuring it’s in the correct array format for AirTable attachments.
Step 5: Implement Review, Approval, and Posting Automation
This final stage ensures you maintain control and automates the publishing process.
1. Review and Approve in AirTable: After the automation generates the post copy and image, these are saved into AirTable. You, as the user, will review the generated content.
If you’re happy, change the status in AirTable to “approved”. If not, you have the flexibility to “make changes to the copy” or “replace the image with an image of your own” directly in AirTable.
2. Scheduled Posting Automation: Set up a separate, simple automation to run on a schedule (e.g., once or twice a day).
This automation should search your AirTable for records with the “approved” status and “LinkedIn” as the social channel.
Conditional Image Handling: If an image exists in the attachment field for an approved post, the system should download that image.
If no image exists, it will simply post the text.
Post to LinkedIn: Connect this automation to your personal LinkedIn account. It will then post the text and the downloaded image (if applicable) to your LinkedIn feed.
Update AirTable: Finally, the system will update the AirTable record to “posted” and record the date it was published.
By following these detailed steps, you can build a robust LinkedIn automation system that provides fresh, engaging content daily without becoming repetitive, and crucially, gives you full oversight and control.