While AI has the power to enhance our work, it cannot replace the essential insights gained from conducting actual UX research and analysis.
AI is a tool that helps us get the job done more efficiently; it does not replace us.
With over 20 years of experience in UX research, I was initially skeptical about integrating AI into our research practices. However, I’ve found that Generative AI (GenAI) tools like ChatGPT, Google Gemini, Microsoft CoPilot, Claude.ai, etc., are immensely helpful. They do more than just assist in analyzing data to uncover insights; they also excel at transforming these insights into effective deliverables such as UX personas and journey maps.There are many ways to write AI prompts to create UX personas and journey maps.
Just google “How to use AI to create UX personas?” and you’ll find plenty of resources (some of which I’ve used and cited at the end). So, why am I writing this article if there’s already a lot out there? After a year of research and experimentation trying out several variations, I’ve discovered a successful approach using a mix of prompts with unique nuances that don’t appear to be covered in anything I’ve read. So, I wanted to share my findings in the hopes that they could also help other fellow UX researchers. As our field continues to evolve, sharing our findings and fostering collaboration will be essential to unlocking the full potential of AI in UX. On that note, I invite you to share your experiences in the comments section if you’ve tried any of these prompts or something different that yielded great results.
Personas
There are two main types of UX personas: those based on actual user research and those based on assumptions. Whether you’re creating data-informed personas or assumptions-based personas (a.k.a. ‘proto-personas’, ‘ad-hoc personas’, ‘alignment personas’, etc.), GenAI can help immensely.
✴️ Side note about proto-personas: In my experience, many people outside the product team don’t understand the difference between personas and proto-personas. Therefore, when creating proto-personas, I prefer to use the term ‘assumptions-based personas’ so it’s clear that research still needs to be conducted. There’s nothing worse than presenting a proto-persona and having stakeholders think they’ve checked the “persona box” and can move on without conducting actual user research!
If you’re interested in learning more about how to create proto-personas, I highly encourage you to check out Tamara Aldin’s great work on alignment personas.
Even when a persona is based on user research, the distilled information is presented in a fictional way while still accurately describing the average user of the product. Whether it’s a name, background story, tagline, etc., these made-up elements help bring a persona to life, making it authentic and memorable while enhancing its ability to resonate with the audience. I don’t know about you, but I always struggle with writing these fictional bits of a persona, and it’s where I get the most help from GenAI.
After conducting research and analysis, I know what data I want to include in a persona and how it should be segmented, but I struggle with the best way to convey it. Some of you may consider this the fun part of persona creation, but personally, I can spend hours just spinning my wheels on coming up with the right tagline. I’m a data nerd and love qualitative analysis, but I don’t necessarily consider myself the most creative writer, which can sometimes make a persona truly meaningful. It’s not to say that I can’t get there alone, but it takes me much longer than someone with this gift. Enter AI!
Recently, I’ve been experimenting with using HeyMarvin, ChatGPT Team, and other GenAI tools to help me fill in the creative writing blanks that I typically struggle with when creating personas. To use AI in this way, I’ve found that it’s most effective if you do the heavy lifting yourself first.
Step 1: Gather all your research data and insights
Collect all your interview transcripts and/or research reports if you’ve conducted user research. If you’re doing proto-personas, you hopefully have a lot of secondary research documents to analyze (brand strategy & industry reports, EOY reports, previous research reports — survey results, market research reports, customer service reports, etc.) or some kind of document(s) that captures the team’s assumptions of the target audience.
Step 2: Decide on your persona sections and behavior prompts
No two personas are created the same. Choosing the sections for your persona that fit your research needs is essential. Every element of a persona should have a purpose, meaning that it should help a team to better understand and empathize with that user group. Always keep the background relevant to the context of your product or service. There’s no point in including excessive personal background information that does not impact how the persona uses the product or its features.
🎦 If you want to learn more about deciding what information to include, watch my free 1-hour course, “How to Create and Use UX Personas”.
Step 3: Upload documents to GenAI tool (i.e. your dataset)
Different tools will allow you to upload different file formats. Currently, you can upload PDFs, TXT, JPEG, PNG, DOCX, CSV, and XLS file formats in ChatGPT Plus or Team.
✴️ Data privacy: Any file you upload to ChatGPT Plus is “retained indefinitely within the service, and those files may also be used by OpenAI to train its models, so it’s best to refrain from uploading files with any important personal information…”. If you require a more secure option where chats are not used for training and data is encrypted, you’ll need to use a ChatGPT Team account or invest in a secure UX research analysis tool like HeyMarvin or Dovetail. I also want to note that there is currently a security issue with ChatGPT Team where it’s impossible to restrict team members from inviting new members to the workspace, so you must review your workspace’s Members page regularly. ChatGPT Enterprise does not have this security weakness. Please check with your IT team before using any GenAI tool with your customer and/or company data to ensure that you adhere to their data security regulations.
Step 4: Provide AI prompt
You can adjust the following prompt below to fit the needs of your persona…
Step 5: Review and refine results
If you find that the first response is not in-depth enough or want it to include specific information, you can ask GenAI to refine it in several ways. For personas, I like to provide behavior prompts. For instance, give the following prompt below to ask GenAI to re-write certain sections of the persona. By the way, these are the same prompts I use with cross-functional teams when facilitating persona workshops to develop proto-personas…
This follow-up prompt should result in a more rich result. You could choose to provide the behavior prompts in your initial GenAI prompt, but I’ve noticed better results if I spoon-feed the data and do iterative refinement. Putting too many demands in one prompt confuses the model. Patrick Neeman, author of Mastering AI Assistants for User Experience Designers and Product Managers, suggests including user research questions in the prompt to “ensure it accurately reflects real users.” He suggests adding to your prompt, “Generate user research questions that would validate this user persona as correct.”
Using large language models (LLMs) to segment a user base into distinct personas based on behavior patterns, preferences, demographic data, etc., is super helpful. However, it is important to remember that you can inadvertently perpetuate or even amplify biases if they’re present in the data you’re using to train them.
✴️ About accuracy: Remember that all GenAI tools are fallible and can provide wrong answers. Ultimately, you must do your due diligence to ensure the information provided accurately reflects your collected data. If the dataset you’re using contains historical biases or imbalances (e.g., in terms of gender, ethnicity, or socioeconomic status), the model’s outputs will likely reflect these biases.
Whether data-informed or assumptions-based, creating UX personas is a nuanced process that significantly benefits from GenAI’s support. These tools can assist in crafting the intricate details that make personas relatable and valuable, particularly for those who may not naturally lean towards creative writing. By meticulously gathering and analyzing user data, defining persona elements that genuinely reflect user needs, and strategically using AI to enhance the creative aspects of persona creation, we can develop more accurate and engaging user personas.
Journey Maps
Photo by UX Indonesia on Unsplash
You would follow the same steps outlined in creating personas for creating journey maps but with some noted differences to Steps 2 and 4.
Step 2 would be to decide on your journey map elements. You want to spend a good deal of time defining the stages of your journey and the user data you want to present during each stage (e.g., actions, thoughts, jobs-to-be-done, etc.). I believe GenAI should not handle this heavy lifting. Instead, you will include this in your AI prompt along with your dataset (i.e., the same docs you uploaded above in Step 3, along with your finalized persona).
For Step 4 (the AI prompt), I like to use a prompt that details:
Who is on the journey (i.e. your persona)
What journey they’re on
The stages of the journey
The information we want to include during each stage (e.g. actions, thoughts, jobs-to-be-done, etc.)
For example, if I were creating a journey map for ordering food through a mobile app, my prompt might look like…
‼️ When reviewing your results, if the journey map does not accurately reflect your dataset, GenAI might be making assumptions!
LLMs are very good at predictive analytics and forecasting user actions, and when they take a guess, it’s not called out in the results unless you specifically instruct them to do so.
For this reason, I prefer to refine the prompt to learn where the gaps in the research are…
💡 TIP: If you want a response to be in a specific format, you should be sure to specify this in your prompt (e.g. “Put the response in a table.”)
✴️ Including opportunities & solutions: For each stage of a journey map, I typically like to provide an additional “swim lane” for Opportunities (suggestions for improvements based on the identified touchpoints, emotions, and frustrations) and Solutions (proposing design or feature enhancements that could improve the UX based on the insights gathered). To me, this is the sweet spot where the research data in the journey map becomes the most actionable. I’ve found that it’s best NOT to use GenAI to help write these lanes. Typically, the dataset doesn’t include the collective team’s expansive knowledge to solve the UX problems presented in each stage.
I’ve found using AI with these prompts can significantly streamline the creation of journey maps, making the process quicker and more efficient. By leveraging AI, we can more easily integrate the findings from complex datasets into each stage of the user journey, ensuring that every aspect, from actions and thoughts to challenges and interactions, is accurately captured. AI’s capability to predict user behavior and generate detailed visualizations can enhance the effectiveness of journey maps. However, it is crucial to manually review and adjust these AI-generated maps to ensure they incorporate human insights and accurately reflect real user experiences to provide a more actionable and comprehensive view of the user journey. While AI can assist in forecasting and visualizing user behavior, the strategic inclusion of human insights remains invaluable in crafting a journey map that truly reflects and improves the user journey.
Moving Forward: Balancing AI Efficiency with Human Insight
In conclusion, as we continue to integrate GenAI into the UX design process, it’s crucial to find the right balance between leveraging its computational power and maintaining the human touch that is essential for creating authentic and impactful user personas and journey maps. By combining AI’s efficiency with our expertise and creativity, we can enhance our workflows to produce deliverables that are profoundly resonant with users’ actual needs and experiences. I encourage my fellow UX professionals to experiment with these tools, share their experiences, and continue to push the boundaries of what we can achieve together in this exciting intersection of technology and design.
If you’re interested in learning more, consider joining the AI for UX Slack group created by Patrick Neeman. We talk a lot about using AI for UX to better understand how to properly use it to advance our industry. I hope to see you there!
Further Reading
Neeman, Patrick. uxGPT — Mastering AI Assistants for User Experience Designers and Product Managers. Available at GPT Prompt Guides.
Aguilar, Nelson. Your Guide to Uploading Files to ChatGPT (and Why You Would Want To). Available at CNET.
Jones, Nate. Your Pocket Guide to Prompt Engineering: How to Get the Most from AI Models. Available at Nate’s Newsletter.
Harley, Aurora. Personas Make Users Memorable for Product Team Members. Available at Nielsen Norman Group.
Adlin, Tamara. Alignment Personas Framework. Available at Adlin Inc.
Mulder, Steve. The User is Always Right: A Practical Guide to Creating and Using Personas for the Web. Available at Amazon.
Morrison, Dylan. UX and AI: Developing a user persona with ChatGPT. Available at Medium.
Bulman, Bill. Crafting Personas with AI-Augmented Research (a step-by-step guide). Available at LinkedIn.
