Day 49 of 100 Days Agentic Engineer Challenge: AI Agent for MarTech

What’s the best market to dive into with AI agents? To start off, marketing and sales seem like a great choice. A simple example would be creating an AI agent to handle social media marketing campaigns. Before we jump into testing that out, let’s take a look at my daily task routine.

Daily Tasks Routine

💪 Physical Activity — I did 70 squats.
😴 Seven Hours of Sleep — I slept for 7 hours.
🤖 AI Agent — I’m working on building a true AI agent, but I’m running into some issues with the social media APIs. I’ll try switching to another provider.
⏳ PAIC — In queue.
📊 Data Science — In queue.

If you want to know what all these tasks are about, read the introduction to the 100 Days Agentic Engineer Challenge.

AI Agent for Social Media Campaign

Every new project should kick off as a Minimum Viable Product (MVP). Let’s identify the essential features needed for an app that an AI agent could use to set up and eventually manage an entire social media campaign for us.

Goal: 100 Leads

Type of marketing: Social Media Organic

Tasks:

• create posts, including text and photos, for an initial 21-day period. After that, optimize and iterate with new content every 7 days
• plan posts with calendar
• autopublish
• Human-in-the-loop approval
• metrics

Tools:

• Use the OpenAI GPT API to generate text for posts and to build an automated review system
• Flux API to create images
• Social media API for automatic publishing of posts and tracking metrics
• LP with integrated Opt-In (MailChimp or GetResponse) for email capture and tracking statistics

The concept is straightforward: the agent starts by using the provided product data to create a customer persona and generate social media posts, each with a link to a landing page (LP) that includes an opt-in form. After the initial 21-day campaign run, the agent reviews engagement metrics and conversion rates to optimize the campaign for the following 7 days, focusing on the best-performing posts. This approach allows for full autonomy and continuous self-improvement. Additionally, we can incorporate a human-in-the-loop at key stages to maintain oversight and control over the agent’s actions.

Status of the AI Agent

I’m having trouble exchanging the code for a token with the X API, and it seems like I’m not the only one facing this issue. It might be related to my local development environment. I’m planning to try a Firebase integration as an alternative. Meanwhile, I’ll also explore creating something similar using Meta’s Threads platform.

✅ FastAPI and Dependencies: Installed
✅ Svelte and Dependencies: Installed
✅ UI — Onboarding Page: Completed (Users can input their expertise or business category and target follower count)
✅ UI — Dashboard Page: Completed (Displays user-provided data and Facebook-related insights)
✅ Integration of News API: Completed
✅ Integration of Azure OpenAI API: Completed
✅ Strategy Generation: Completed
⚠️ Integration of X.com API: In progress (Authorization done, now trying to get token)
⚠️ Integration of Threads API: In progress
🔄 Firebase Integration
🔄 Content Strategy Development: In progress (I’ve tested the Phidata framework and will now continue with PydanticAI)
🔄 Testing ReAct Agents: In progress
🔄 Studying the PydanticAI and Phidata docs