When we create assistants for various use cases it helps to:
- Know how to communicate our thoughts, ideas and tasks
- Have a rich vocabulary
- Structure our communication in prompts format
Today, I want to focus on prompts - How can we write a prompt for anything?
- Define what you want to do.
To write a prompt, we need to know what we want it to do, right?
AI has a lot of data coming from various areas and sources. As a result, it can give answers and help with whatever task we give it. The other side of this is generic output. Because of the wide range of information, if we’re not specific, it will give us a basic overview of the information.
When choosing our use case, the more niche we go, the more specialized prompt we can write and consequentially the better our assistant output.
2. Use AI to help write prompts
Who better to format our prompts than AI, right?
It’s who we’re communicating with after all. We can go straight to the source and ask how to structure prompts.
Starting Prompt:
Hey.
I am looking to create assistants with AI large language models.
I need help with “project organization, scheduling, youtube videos drafts, email drafts.”
As a large language model yourself, what do you think would be the best way to structure this?
I am looking to have an assistant for each area I need help with, or if possible have some of the areas combined.
Could you also suggest what my instructions to the AI assistants would be.
Thank you very much.
Further interaction:
Can you write a core prompt for “use case” please?
I want to create an assistant that would help me make “youtube videos.”
“I have voice transcripts. I want to turn those into video scripts for youtube videos.”
What do you think would be the best way to do it?
I want to use LLM to help me do this process.
What would be the prompt for the llm?
3. Gather information online
People share a lot of information online. A lot of step by step guides, frameworks, processes.
When you find something that feels useful to you, consider if you can turn it into a prompt for an AI assistant.
An example of a copywriting framework I received in a newsletter from Justin Welsh, which I turned into a prompt:
I would like you to help me write LinkedIn posts.
Follow this framework:
Problem: Identify your audience's pain points and challenges.
Amplify: Amplify the consequences of not addressing the problem.
Story: Share a relatable story or example that illustrates the problem.
Transformation: Offer a solution that transforms the situation.
Offer: Present your product or service as the key to achieving the transformation.
Response: End with a clear call-to-action that encourages people to take the next step.
4. Always review with a Critical Mind.
Whatever we create, it’s important to communicate our values and personality. We can prompt and fine-tune AI as much as we want, but in the end, it is not us. It does not have our experiences, feelings, dreams, etc.
Reviewing the output and making sure we feel comfortable sharing this with others is an irreplaceable step in working with AI.