Marketing > Insightologist Insightologist® Insightologist.com > Gain Insight - How 2 chatbots co-create / improve prompts

Gain Insight - How 2 chatbots co-create / improve prompts

Watch two AI engineers improve (with rationale) one of your chatGPT prompts in a dialog. Start with this prompt before providing the prompt you'd like to improve.

What to use it for and what to expect

You ask chatGPT to act as two top ai engineers and with this expertise, to perfect an existing prompt for you. They will go through a dialogue and explain all suggestions and the creation process along the way. You can interfere or interact at any time.

Preparation - what you need to get started

  • chatGPT (I used GPT-4 August 3 Version) or Bard or similar LLM (large language model - a machine learning term) chat bot.

  • A basic understanding of ai chat bot prompting.

  • An existing prompt you want to improve.

  • Plenty of playtime - this is addictive.

Initial prompt to perfect any prompt - together with chatGPT

In a minute I will provide a prompt that I need you to perfect as follows: Act as the two best and most experienced ai engineers, data scientists, machine learning experts. Develop and prefect a ChatGPT-4 prompt in a dialogue with each other that I will provide to you in the next step. Go through several iterations of improvements and explain the development and improvement process. Deliver the perfect prompt at the end. The prompt needs to include an option to spread over several response sessions, so the response to that prompt will not be limited. Make this function user friendly. In terms of the prompt that I will deliver in a minute: I need to use this prompt universally for many different topics.

chatGPT response

Understood. Please provide the initial prompt you have in mind, and we'll work together to refine and perfect it using a collaborative approach.

What happens next?

Paste your prompt into the text field of chatGPT and watch a dialogue unfold.

I chose a prompt to create user personas and user journey for a website.

Here is what happened next:

AI Engineer A: Alright, we've got our initial prompt. Let's start by breaking it down. The user wants a detailed user persona for a virucidal coating application on glass.

AI Engineer B: Right. It seems the core structure of the request is about creating user personas and journey maps for specific topics. There's a lot to unpack there, but it's crucial that we provide a dynamic structure to allow this prompt to be universally applicable.

AI Engineer A: Exactly. First off, it's a two-parter. The user wants both a persona and a user journey. Let's start with the user persona. We have placeholders for demographics, motivations, goals, and pain points. We should ensure the structure supports the addition or removal of such categories without breaking.

AI Engineer B: For the universality aspect, we should look at making the placeholders within brackets more dynamic, so that any topic, demographic or any other attribute can be replaced. Let's refine that:

Revised User Persona Prompt:

"Create a detailed user persona for a [Topic/Product]. Include attributes such as [Attribute 1, Attribute 2, ...], motivations [e.g., Motivation 1, Motivation 2, ...], goals [e.g., Goal 1, Goal 2, ...], and challenges [e.g., Challenge 1, Challenge 2, ...]."

AI Engineer A: That looks cleaner. Now, let's tackle the user journey. The structure should allow for touchpoints, emotions, and actions to be clearly defined, and with the added challenge of spanning multiple response sessions.

AI Engineer B: Right, and we should also make sure that the recommendations are detailed and tie back to both the persona and the journey map.

Revised User Journey Prompt:

"Map out a detailed user journey for someone using/interacting with [Topic/Product]. Include touchpoints like [Touchpoint 1, Touchpoint 2, ...], emotions they might experience [e.g., Emotion 1, Emotion 2, ...], and actions they would likely take [e.g., Action 1, Action 2, ...]. Respond with a detailed analysis and in-depth recommendations to cater to the persona at every step, spread over [X number of sessions]."

AI Engineer A: Nice. We've clarified the journey map and added an option for multiple sessions. Now, to address the universality, the prompt's placeholders should be easy for the user to understand and replace, ensuring they know exactly what information needs to be filled in.

AI Engineer B: Agreed. Let's summarize and integrate the two parts into a final prompt, highlighting that it should be adjusted as per the user's requirements.

Final Universal Prompt:

"Create a detailed user persona for a [Topic/Product]. Include attributes such as [Attribute 1, Attribute 2, ...], motivations [e.g., Motivation 1, Motivation 2, ...], goals [e.g., Goal 1, Goal 2, ...], and challenges [e.g., Challenge 1, Challenge 2, ...].

Map out a detailed user journey for someone using/interacting with [Topic/Product]. Include touchpoints like [Touchpoint 1, Touchpoint 2, ...], emotions they might experience [e.g., Emotion 1, Emotion 2, ...], and actions they would likely take [e.g., Action 1, Action 2, ...]. Respond with a detailed analysis and in-depth recommendations to cater to the persona at every step, spread over [X number of sessions]. Adjust the placeholders within brackets to fit the specific topic or need."

AI Engineer A: This feels much more structured, cleaner, and dynamic. I believe we've achieved a universally adaptable prompt, with the flexibility the user was looking for.

AI Engineer B: Agreed. The user can now adapt it to any topic, ensuring they get comprehensive results each time.

Marketing > Insightologist Insightologist® Insightologist.com > Gain Insight - How 2 chatbots co-create / improve prompts