Happy Tuesday!

This past week, I automated the generation of social media post drafts using Make.com and AI. Here is how I did it:

Current State - Drafting Manually

First, I want to talk about how I manually manage my social media posts. There are three categories of posts I typically share on social media:

1. The Problem Solver - This is where I identify a problem my audience has and offer solutions

2. Motivational - This is where I post about topics I am struggling with and ways that I stay motivated

3. What I Learned - This is where I share something I have learned about by explaining a topic or creating a brief tutorial

Systemize the Process

As a first step towards automating this process, I needed to understand how I could systemize my process. Here is what I came up with:

1. Who is my audience?

2. What is the goal this audience is going to achieve?

3. List out 5-10 ideas about this goal:

   1. Problem Solver - identify problems

   2. Motivational - Identify tools to stay motivated

   3. Learning - Identify topics that might be useful to teach my audience

4. Pick the top 3 from these lists and write a draft posted using pre-defined formats

   1. Problem Solver: {Are you having this problem?} {Do do this} {Instead do this}

   2. Motivational: {Lately I've been struggling with X}{This is how I stay motivated: {list of tools}}

   3. Learning: {What is X?} {Longform essay or tutorial} {Image or video}

Automation with Make.com

With a system in place, I was ready to start tackling automation. I decided to attempt automating this process using Make.com, a 'No-Code' website that allows users to automate tasks and workflows. I first learned about Make after watching a video on YouTube from Nick Saraev called "Make.com Email System & AI Autoresponders." In this video, Nick explains how to connect to an email inbox, get the data from emails with a specific subject, then connect OpenAI GPTs and create completion to generate an email response based on the data provided in the email. While I do not need to create email auto-responders, I did feel GPT completions could help speed up my social media post-generation process.

Prompt Engineering

With this in mind, I got to work. Make.com makes it simple to connect to a lot of tools, including the Google suite of cloud applications. Using these would be the easiest. I started by creating a Google form that would capture steps 1 and 2 of my system, identifying my audience and their goal. I set up an automation in Make that watched for responses to that form and captured the form data. Next, I needed to use that information to create a prompt for the GPT completion. For this, I had to learn about prompt engineering.

What is prompt Engineering? According to this article from Masters of Code, "GPT prompt engineering is the practice of strategically constructing prompts to guide the behavior of GPT language models, such as GPT-3, GPT-3.5-Turbo or GPT-4. It involves composing prompts in a way that will influence the model to generate your desired responses.” The article explains that a great prompt is composed of four items:

1. Context - Brief background information to set the context of the conversation

2. Instructions - Clearly state what you want the AI Model to do

3. Input Data - Give specific examples for the model to build upon

4. Output Indicator - State the format you want the response in

With this in mind, I created a prompt to assist in generating the "Problem Solver" post type:

• Context: You are an intelligent social media manager, generating social media post ideas for the user.

• Instructions: You're monitoring a form submission that receives ideas for a problem the user wants to address for their audience. Your job is to digest the input response and produce a simple list of the issues the audience faces in achieving a specified goal. Ensure the responses are spartan and do not use frilly language. Use 17 words or less.

• Input Data: I want to teach my audience, busy people working full time, how to achieve their goal: make time for continuous learning. What are the ten biggest mistakes they make when trying to make time for learning?

• Output Indicator: Sure, here's a list of 10 common mistakes busy professionals make when trying to make time for learning:

  1. Not prioritizing learning.

  2. Ignoring small pockets of time.

  3. Multitasking while learning.

  4. Setting unrealistic goals.

  5. Skipping breaks.

  6. Not using technology.

  7. Lack of planning.

  8. Fear of missing out.

  9. Not seeking help.

  10. Waiting for perfect conditions.

To get the best results possible, I created two additional prompts, one to take each item in that list and come up with a "good" and a "bad" for each mistake.

Once the OpenAI GPT generated this list, my last prompt took the new list of "good" and "bad" and generated sample posts in the following format:

Once the automation ran through the prompts, my automation in Make.com created a Google Doc in my Drive folder with the draft posts. All I have to do now is review, edit, and post!

This week, I learned to automate my social media posts with Make.com and AI. I'm excited to learn more about AI in the next few weeks!

Tuesday Top Three

This is the section where I share three things I enjoyed and/or learn from in the past week.

Youtube Video

A Step By Step Guide To Making $10,000 a Month - Ali Abdaal

Book

Feel Good Productivity - Ali Abdaal

Podcast

Building in public on Twitter, with Kevon Cheung - Digital Marketing Troop

Let me know if you found value in the week's newsletter! Enjoy the rest of your week!

-- Summer