Skip to content
All posts

How to build a Social Media AI Agent

Picture of Will Sullivan By   ·  3 minute read

Key Takeaway: Tell a generative AI chatbot your problem and to draft a solution.

The Business Challenge

Maintaining a consistent social media presence for my firm while helping clients mature their RevOps and business intelligence capabilities.
 

My goal: to spend only an hour a day on social media & marketing so I can focus on clients, AI product development, and the latest SaaS industry trends & frameworks.  

Perspective: Having been in combat, I know plans are not enough.  You need to prepare for chaos.

  • In the army, that meant establishing standard operating procedures (SOPs).

  • In the corporate world, that means establishing repeatable systems.

  • In the era of generative AI, it means creating an AI Agent to build and manage a repeatable system for you.  

     

Executive Summary

 Key Definition: AI Agents are intelligent systems that execute multi-step sequences for task completion, and are capable of collaborating with other agents to achieve complex goals.

Business Value Created: Additional two hours per day to focus on clients. Reducing Social Media Marketing time requirement from 3 hours per day to just 30 minutes.


Action Steps: 

  1. Brainstorm (with AI)

  2. Write a Draft (with AI)

  3. Test & refine (Human + AI)

  4. Deploy instructions to AI agent tool (Human)

  5. Iterate & refine (Human + AI)

Output: Multi-week Content Calendar for Substack, LinkedIn and X:

Next Steps: Build out a human-in-the-loop agentic marketing content creation process.

The Build Process

 1. Initial Brainstorming (5 minutes)
 
I started by telling Gemini (I used Gemini Advanced to test out its export to sheets functionality) my business challenge then asked it to write a prompt to create an agent to help me. The key was providing specific needs and steps for an agent to create a content calendar that could:

  • Receive a theme or blog title

  • Generate platform-specific recommendations

  • Create exportable scheduling documents

Prompt 1: Brainstorming with Gemini

2. Core Development (5 minutes)

 The development phase focused on prompt engineering, where I:

  • Created initial prompts for blog post analysis

  • Developed table structures for content calendars

  • Implemented markdown formatting for easy export to Google Sheets

  • Added platform-specific content tailoring instructions

Initial response from the agent

Gemini provided me with a typical conversation text output listing days, blog titles & content recommendations, however, I wanted a calendar view.

Prompt 2: Create an exportable content calendar

Definition: Markdown is an open source code language that uses characters found on a  keyboard to create format changes rather than clicking on a button in Microsoft Word or Google Sheets enabling easier formatting for websites.

Example: Double asterisk before or after a word or phase makes it bold.

  **This would be Bold in Markdown**

In my prompt, I used the | symbol which is commonly referred to as the “pipe symbol” to designate columns and provides a simple table structure. 

Platform  |  Monday  |  Tuesday  |  Wednesday  |  Thursday  |  Friday

LinkedIn  | put blog title and content in the respective column for the recommended day to posts

Gemini responded with a Content Calendar in a table format and had the “Export to Sheets” action button at the end of its response which I of course used.

Raw export to sheets before formatting

3. Testing & Refinement (5 minutes)

 

Testing revealed several key improvements needed:

  • Multi-week calendar capability

  • Platform-specific content requirements

  • Manual formatting of the output

Formatted Content calendar

Prompt #3: Prompt to create a multi-week calendar 

Step 4. Deploy the AI Agent (5 minutes)

 

In Gemini, I clicked on Gemini’s Gem manager in the left side panel.

I clicked “create” and pasted my instructions into the Gem.

Click Save and a new chat will appear where you can interact with your new AI Agent.

To edit an agent: click Gem Manager in the left side panel and click the pencil icon next to my agent.

Real-World Application

 The resulting AI Agent now:

  • Creates comprehensive multi-week content calendars

  • Generates platform-optimized post suggestions & schedules

  • Seamlessly exports to Google Sheets

  • Allows me to focus on analyzing & refining rather than coming up with dozens of social media ideas

Three blog post ideas can now be transformed into 36+ social posts across:

  • Substack newsletters for long-form content

  • A series of LinkedIn posts for my professional network

  • X threads that break down key concepts

This is just the start to my human-in-the-loop agentic workflow.

Human-in-the-loop agentic marketing approach:

  1. Social Media Manager Agent creates a Content Calendar

  2. Human reviews & refines

  3. Copy Write Agent writes Social Media Posts

  4. Human reviews & refines

  5. SEO agent develops a Key Word Strategy

  6. Human reviews & refines

  7. Social Media Manager Agent integrates keyword strategy into copy write 

  8. Human reviews & refines

  9. Marketing Agent publishes to marketing platforms 

 

If you are interested in learning more about AI Agents, or our approach to accelerating RevOps and Business Intelligence capabilities schedule some time here.