5 Step Framework for Prompt Engineering
- Task - what you want the AI to do
- Persona - give the AI a role to embody for more specific results
- Format of output - request something more structured, such as a table
- Context - the more context you provide, the better the output
- References - provide examples that would be included
- Evaluate - after getting the output, really look at it and determine if this is what you were looking for
- Check for hallucinations
- Check for bias - AI is trained on human created content, and comes with biases of race, gender, and more
- Iterate - prompting is rarely done after the first search. Always add more detail and try again!
- Revisit the promoting framework, go through the 5 steps again
- Separate the prompt into shorter sentences, feed to the AI slowly
- Try different phrasing to make output more compelling (example, instead of asking for a marketing plan, ask to describe how the product fits into the lives of customers)
- Introduce constraints
Prompt Chaining
Chaining guides GenAI through interconnected prompts to add complexity. Example: Combine the previous three suggestions with a focus on _____. Find the catchiest and most impactful combination.
- Chain of thought prompting - ask the AI to explain it's reasoning as a step by step process. Let's you understand more and prompt further.
- Tree of thought prompting - lets you explore multiple paths simultaneously and brainstorm different options.
