How LLMs help can help you draft better personas.
In this guest post Ze'ev Abrams share his personal prompts to iterate on user personas.
In this guest post Ze'ev Abrams, founder of Iteraite.com shares his personal techniques to iterate on user personas with AI.
TL;DR ✨
✅ Using AI for persona creation can help identify potential product issues early in development.
✅ LLMs can create more diverse user personas than “the ideal one”, such as “less-ideal” or “skeptical” user personas.
✅ The article provides a prompt template for creating AI-generated personas with ChatGPT or Claude.
✅ You will also find a free app you can use to quickly generate personas.
Using AI in your work as a Product Manager should be something you do on a daily basis.
By “AI”, we of course mean current Large Language Models (LLMs), such as ChatGPT or Claude, which can help you write things more efficiently, but also provide you with variations that help you better think about your product.
Let’s take an example - that of creating a Persona.
The “correct” way to generate a Persona is to collect many interviews with actual customers, and then amalgamate the best ones into a single Persona that is the quintessential example of who you’re building the product for.
But how real is that Persona?
Is it your ideal customer, or your average one?
The advantage of LLMs is that you can both generate that ideal customer, as well as help you think about other customers who may not fit that ideal customer profile.
This is the case with many other questions you may want a “second opinion” on - and an LLM is a great provider of second opinions!
This post will briefly provide some tips for creating better, and more interesting Personas, and also provide links to a free web applet that can create these personas quickly, as well as open-sourcing the code for the applet.
Prompt Basics: Give it Context, Like a New Intern.
So, what is the technique you should use to ask the LLM these questions?
LLMs aren’t magical - they base the next word of their output off of the previous combination of words.
Giving an LLM short, direct and explicit information and instructions will improve its result, whereas having long chats with extraneous information before you ask it what you want will make it lose its way (and potentially “hallucinate” - where it randomly selects output based off of its extensive training information, instead of off the information you provided it).
It’s not a matter of more is better, as rather, contextually specific is better.
This means that you should provide an LLM with:
The proper Context,
Ask it to do something Specific
With Instructions on how to do it will provide better results.
Fast and Precise Personas Using LLMs.
Let’s try this with an example: Imagine you’re creating a product that uses AI to help write blog posts for Product Managers.
You need to provide it with
Who this is for,
What problem this is solving,
What it is you’re trying to do,
And then ask it to define your Persona.
If this sounds familiar to your work as a PM - then you are correct! You use the same “common sense” tools in your everyday work as a PM as you would with interacting with an LLM!
Here’s the context section of our prompt:
You are a Product Manager working on a new product.
You are working on the following project/product: An application that helps Product Managers write blog posts for their company using AI.
The Problem that this product solves is:
Product Managers don’t have time to write long blog posts. Many Product Managers aren’t good at writing. Some Product Managers don’t use English as a primary language.
The Target Customers this is for are:
Product Managers at large Enterprises, Freelance Product Managers looking to increase their influence
Now, ask it a specific question, and provide it with Instructions on how the output should look:
Write a User Persona for this product. Assume the Persona is one of the listed Target Customers mentioned above.
For the persona, make sure to list:
Their name1. Overview
2. Goals
3. Behaviors
4. Pains
5. Needs
For each of the 5 categories, write 3 short 1-line bullets.
Copy both sections into your LLM chat, press <Enter> on your LLM, and you’ll get a nice response!
You can change the instructions if you like something different (like their hobbies, if it’s important).
All you need to do now is to take your own project/product, replace the data in the Context section (while remembering to keep the text before the colon - for context!), and you have your Persona!
Play with LLMs to go deeper.
OK - but we can do better - let’s ask it to generate Personas that are not our ideal persona. That’s just a copy-paste away!
All you need to do is replace that first bullet in the Specific section above. Here are 2 examples:
Write a User Persona for this product. Assume the Persona is one of the listed target customers mentioned above, however, assume that it is a less-ideal customer who is less likely to like your product, and it doesn't fit their needs 100%.
Or,
Write a User Persona for this product. Assume the Persona is one of the listed target customers mentioned above, however, assume that it is someone who is not at all likely to like your product, as they are skeptical, and it doesn't fit their needs and pains.
These prompts will give you “less than ideal” Personas, which are actually more likely for you to come across in reality. They may help you quickly find problems and issues with your product that you may not have thought of, and that you may want to address even before building your product!
You can recreate these types of results by copying and modifying the prompts above, or by crafting a similar style to fit your own requirements.
The fundamentals are the same though - always keep precise context and instructions. You can do this with any free LLM, or you can use this free web-applet I created that follows the instructions provided here.
You can access the applet here - it will produce the 3 Personas as above.
If you are more technically savvy, I have open-sourced the code to produce these results here.
If you like this type of template, you can find others like it at our full web-app here, or through iteraite.com.
A few words about The Product Courier
👋 We’re Lucas Nilsson and Toni Dos Santos, Co-founders of The Product Courier.
We help Product Professionals leverage the game-changing potential of AI and become better at their jobs.
We’ve recently shared our personal library of 30+ prompts for ChatGPT and Claude 3.5.
All our prompts are organised by use cases (market research, positioning, data analysis) and help you get great results from GenAI.
As a Product Courier reader, enjoy 30% off with code "PRODUCTCOURIERCOMMUNITY30".
If you enjoy reading this article hit the ❤️.
It helps us get better reach and encourages us to keep sharing great content from the community.
See you next Wednesday!