While there’s no shortage of content on how AI is changing our jobs and the future of work, it can be hard to figure out where to start. At Manual Labor, we’re all about finding the most efficient and effective ways to work, and sometimes that includes using AI to support. Here’s what we’ve been focused on over the last quarter. 

Note: we have used ChatGPT4o as the LLM in these examples. 

1. Practice Prompt Engineering

No surprises here: above all else, we’ve found it the most helpful to continue experimenting with our prompt engineering. To misquote Wayne Gretzky, “you miss 100% of the outputs you don’t refine.” Or said in a clearer way: “your ChatGPT output has the ability to be stronger as you refine what you put in.” Whether it’s experimenting with which additional details to add to a prompt, setting up a persona, or specifying the desired output length, we have found so many ways to continue pushing what we initially thought was the best result due to prompt refinement. 

One example for data analysis: using ChatGPT4o, we wanted to understand if there were any correlations between two variables in a data set. We pushed past the initial prompt of “based on this CSV, what are correlations in the data?” and continued to provide specificity in ideal output. Points included:

• Setting up the persona (“Pretend you’re an expert data analyst”)
• Explaining the context (“You’re reviewing a data set with [types of variables]”)
• Share what you need assistance with (“I’d like to see the relationship between X and Y variables”)
• Ask for clear insights and next steps (“Based on the available data, what are takeaways so far? What are the actionable next steps?) 
• Remind ChatGPT4o to be specific (“Please be very explicit and call out any additional details you may need”).

‍

Using a marketing campaign example, asking ChatGPT to “Create a marketing plan” will render results, but at a very basic level. Adding further details – like client type, what you’re trying to accomplish, any constraints (time, budget, resources) – will strengthen the output. 

In this case, the updated prompt would be: “Create a marketing plan for a B2C, eCommerce retail client who wants to increase engagement on its homepage. This plan needs to be executed in one month. The budget is $5,000, and there is a full-stack web development team on the project.” 

Additional Resources: For an expansive list on prompt engineering, check out OpenAI’s guide here.

‍

2. Experimenting with different output formats 

For certain use cases, like building out a launch plan, asking ChatGPT to “provide a response in a tabular format” makes for easier scannability. We’ve also loved using markdown format (check out point 6 below) and designing flows and processes using Lucid.  

‍

3. Identifying blind spots 

Identifying gaps in our thinking is a crucial aspect of decision-making. One way we test our ideas is by adding the following line to our prompts: “Please ask questions to gather the information you need. Do not offer a response until you have the answers.” ‍

‍

4. Discovering useful formulas for Google Sheets

In a perfect world*, (Read: when Alphabet inevitably rolls this into an upcoming feature release) there is a toolbar shortcut for every use case in Google Sheets. Until then, we’ve found it helpful to use ChatGPT for quick results, from finding formulas to tips on conditional formatting. What we appreciate the most here: having context on the function, and not just the answer.

An example of a prompt from our operations activity: 

Output: You can use the IF function in Google Sheets to achieve this. Here's the formula you can use in cell F2:

‍

5. Finding new ways to visualize data 

Reviewing a set of data has often confronted us with the question of the best way to visualize it. ChatGPT can create a variety of graphs and charts to suit different data types and highlight patterns. In our prompts, we like to include what we are seeking to accomplish + example correlations + questions on visualizations available.  For example: 

In this prompt’s output, ChatGPT4o provides several visualization options, from scatter plots to heat maps and bubble charts. As a follow-up step, upload your data as a CSV and specify what visualization you’re most interested in.

Additional Resources: For more assistance on data visualization, use this mega prompt from ChatGPT. 

‍

6. Creating mind maps 

We’ve found creating mind maps in ChatGPT a helpful tool when we have initial content ideas (e.g. writing this blog post) and want to find natural extensions of where the content can go. Shout out to DTC Marketing for this idea!

Here’s an example of one we ran for our own content marketing. First, we input a prompt in ChatGPT to create the markdown output:

‍

Output from ChatGPT as Markdown:

## Short Form Content
- Social Media Posts
  - Highlighting AI applications in Manual Labor's workflow
  - Quotes from employees about AI efficiency
- Blog Post Titles
  - "5 Ways AI Boosts Efficiency at Manual Labor"
  - "The Future of Work: How AI is Revolutionizing Manual Labor"
- Email Newsletter
  - Teaser about AI implementation at Manual Labor
  - Link to blog post discussing AI applications

## Long Form Content
- Whitepaper
  - In-depth analysis of AI integration at Manual Labor
  - Case studies illustrating AI's impact on productivity
- Webinar
  - Panel discussion with Manual Labor executives on AI adoption
  - Q&A session with audience about AI implementation challenges and successes
- Podcast Episode
  - Interview with Manual Labor's AI team about their strategies
  - Stories from employees on how AI has transformed their work processes

## Visual Content
- Infographics
  - Visual representation of AI's role in Manual Labor's workflow
  - Comparison of manual vs AI-assisted tasks
- Video Series
  - Behind-the-scenes look at AI technology in action at Manual Labor
  - Testimonials from employees about AI benefits

## Interactive Content
- Interactive Website Feature
  - AI-powered chatbot providing information about Manual Labor's AI initiatives
- Quizzes and Polls
  - Engagement tools to gauge audience understanding and interest in AI integration

‍

Next, we copied and pasted the full markdown from ChatGPT into Markup https://markmap.js.org/.

While any of these ideas may not be complete in entirety, the mind map is a great visual reference to see different types of content at a glance. It also provides inspiration on both content types we may not have originally considered – like launching a poll to our readers to understand their interest in AI – and starting points for blog titles when we feel like we’re stuck!

‍