This AI Agent Manages Your ENTIRE Linkedin Strategy (No-Code)

LinkedIn Automation Agent: No-Code Content Creation

Key Concepts:

• No-Code Agent: An AI-powered system built without traditional coding, using platforms like Relevance AI and Make.com.
• LinkedIn Automation: Streamlining content creation and outreach on LinkedIn using AI agents.
• RAG (Retrieval-Augmented Generation): A technique where a language model retrieves information from a knowledge base to improve its responses.
• Fine-Tuned Language Models: Customized AI models trained on specific data (e.g., past LinkedIn posts) to generate content with a desired tone and style.
• Real-Time Data Access: Providing the agent with up-to-date information through web scraping and database integration.
• Co-Pilot Agent: An AI system that works collaboratively with a human, assisting in tasks rather than fully automating them.

1. Agent Overview and Benefits

• The agent automates the LinkedIn content creation process, saving time on research, ideation, and post writing.
• It provides real-time access to top-performing posts in a specific niche.
• It uses RAG to find similar posts based on topics.
• It drafts posts based on ideas, YouTube videos, or other content.
• It writes posts in a LinkedIn-friendly tone using fine-tuned language models.
• It adds posts to a LinkedIn content calendar or directly posts them to LinkedIn.

2. Agent in Action: Use Cases

• Identifying Top-Performing Posts: The agent can query a database of scraped LinkedIn posts to find the most liked or commented posts within a specific niche over a given time period.
  • Example: "Check the top-performing posts in my niche for the last month."
  • The agent retrieves posts with metrics like likes and comments, allowing the user to identify trending topics and content styles.
• Finding Similar Posts Using RAG: The agent can use RAG to find posts similar to a given idea or topic.
  • Example: "Find similar posts on AI for personalized outreach."
  • The agent queries a knowledge base of vectorized LinkedIn posts to identify relevant content.
• Drafting Posts with Fine-Tuned Models: The agent can generate multiple variations of a LinkedIn post based on a given idea, using a fine-tuned language model trained on the user's past posts and the posts of other creators.
  • Example: "Write a post on human-in-the-loop steps are vital to making AI agents more reliable."
  • The agent generates several variations with a LinkedIn-friendly tone and style.
• Content Calendar Integration: The agent can add drafted posts to a content calendar (e.g., in Notion) for review and scheduling.
  • The user can make adjustments to the post before adding it to the calendar.
• Content Repurposing: The agent can repurpose content from other sources, such as YouTube videos, into LinkedIn posts.
  • The agent transcribes the video and generates post variations based on the transcript.
• Direct Posting to LinkedIn: The agent can directly post content to LinkedIn, although the user recommends using the content calendar for better control.

3. System Architecture and Components

• Relevance AI: The primary platform for building the agent and managing the knowledge base.
• Make.com (formerly Integromat): Used for integrations with external services like AirTable, Notion, and LinkedIn, as well as for running fine-tuned language models.
• LinkedIn Agent: The core component that interacts with the user and orchestrates the various tools.
• Tools:
  • Add to Content Calendar: Adds written LinkedIn content to a Notion content calendar.
  • Post to LinkedIn: Directly posts content to LinkedIn.
  • Write LinkedIn Post: Generates LinkedIn posts using fine-tuned language models.
  • Query LinkedIn Database: Retrieves data from an AirTable database of LinkedIn posts.
  • Get Current Date: Retrieves the current date for use in queries.
• Databases:
  • LinkedIn Post AirTable Database: Stores LinkedIn posts with metadata like creator, URL, likes, comments, and date.
  • LinkedIn Post Knowledge Base (Relevance AI): A vectorized knowledge base of LinkedIn posts used for RAG.
• LinkedIn Scraper: A tool that automatically scrapes LinkedIn posts and updates the databases daily.

4. Detailed Breakdown of Key Components

4.1. Query LinkedIn Database Tool

• Purpose: Retrieves records from the AirTable database of LinkedIn posts.
• Inputs:
  • amount_of_post_retrieve: Number of posts to retrieve (default: 5).
  • sort_by: Sort criteria (likes, comments, or none).
  • filter: AirTable formula for filtering the data.
• Process:
  1. The agent receives the query from the user.
  2. It extracts the amount_of_post_retrieve, sort_by, and filter parameters.
  3. It sends these parameters to Make.com via an API step.
  4. Make.com uses the AirTable integration to query the database based on the provided parameters.
  5. The retrieved records are sent back to Relevance AI via a web hook.
• Example Filter Formula: IS_AFTER(Date, DATEADD(TODAY(), -7, 'days')) (retrieves posts from the last week).

4.2. Find Similar Posts Tool (RAG)

• Purpose: Finds LinkedIn posts similar to a given idea using RAG.
• Inputs:
  • idea: The idea or topic to find similar posts for.
• Process:
  1. The agent receives the idea from the user.
  2. It uses the Knowledge Search step in Relevance AI to query the LinkedIn Post Knowledge Base.
  3. The Knowledge Search step performs a vector search to find the most similar posts.
  4. The retrieved posts are returned to the agent.
• Knowledge Base Configuration:
  • The LinkedIn Post column is vectorized for semantic search.
  • The Vector search type is used for finding similar posts based on meaning.

4.3. LinkedIn Scraper Tool

• Purpose: Scrapes LinkedIn posts from specified profiles and updates the AirTable database and Relevance AI knowledge base.
• Inputs:
  • LinkedIn URLs: A list of LinkedIn profile URLs to scrape.
• Process:
  1. The tool iterates through the list of LinkedIn profiles.
  2. It scrapes the LinkedIn posts from each profile.
  3. It filters the posts to identify new ones.
  4. It transforms the data into a format suitable for AirTable and Relevance AI.
  5. It inserts the new posts into the AirTable database and the Relevance AI knowledge base.
• Automation: The scraper is triggered daily using Make.com, which reads the list of LinkedIn URLs from a Google Sheet.

4.4. Fine-Tuned LinkedIn Post Writer Tool

• Purpose: Generates variations of LinkedIn posts using fine-tuned language models.
• Inputs:
  • type_of_post: Type of content (idea, YouTube, video).
  • idea: The idea for the post (if type_of_post is idea).
  • YouTube_link: The YouTube video link (if type_of_post is YouTube).
  • video_link: The video URL (if type_of_post is video).
  • free_template_CTA_link: Optional link to include in the post.
• Process:
  1. The agent receives the input parameters.
  2. It sends the parameters to Make.com via an API step.
  3. Make.com uses the OpenAI integration to generate post variations using the specified fine-tuned language model.
  4. The generated variations are sent back to Relevance AI.
• Fine-Tuning: The language model is fine-tuned on the user's past LinkedIn posts and the posts of other creators to capture the desired tone and style.
• Make.com Flow:
  • The flow uses a router to handle different content types (idea, YouTube, video).
  • For YouTube videos, the flow transcribes the video before generating the post variations.
  • The flow uses the OpenAI integration with a custom model name to access the fine-tuned language model.

4.5. Update LinkedIn Content Calendar Tool

• Purpose: Adds drafted posts to a Notion content calendar.
• Inputs:
  • title: Three-word summary of the post.
  • text_for_the_LinkedIn_post: The entire LinkedIn post text.
  • LinkedIn_post_type: Type of post.
  • image: Image URL (optional).
• Process:
  1. The agent receives the input parameters.
  2. It sends the parameters to Make.com via an API step.
  3. Make.com uses the Notion integration to create a new entry in the content calendar database.
  4. The entry includes the title, text, and image (if provided).

4.6. Post to LinkedIn Tool

• Purpose: Directly posts content to LinkedIn.
• Inputs:
  • type_of_post: Type of content (video, image, text).
  • text: The text of the post.
  • video: Video URL (if type_of_post is video).
  • image: Image URL (if type_of_post is image).
• Process:
  1. The agent receives the input parameters.
  2. It sends the parameters to Make.com via an API step.
  3. Make.com uses the LinkedIn integration to post the content to LinkedIn.

5. Agent Setup and Prompt Engineering

• Agent Prompt: The agent prompt defines the role, objective, context, and instructions for the agent.
• Role: Expert LinkedIn assistant specializing in ideation and post writing.
• Objective: Ideation, content creation, and publishing to the content calendar or LinkedIn.
• Instructions: Clear instructions on how to use the available tools and when to use them.
• Examples: Provide examples of real-world scenarios to guide the agent's behavior.
• Tool Descriptions: Provide detailed descriptions of each tool, including its purpose, inputs, and how to use it.
• Language Model: Use a powerful language model like GPT-4 or Claude for best results.

6. Conclusion

The LinkedIn automation agent streamlines the content creation process by providing real-time data access, RAG capabilities, fine-tuned language models, and integrations with external services. By combining these technologies, the agent empowers users to create high-quality, engaging content more efficiently and effectively. The key takeaways are the importance of real-time data, the power of RAG and fine-tuned models, and the flexibility of no-code platforms for building custom AI solutions.