In looking for ways to test out AI for real-life work, I posed myself the question: how AI can prove useful in setting up user research, make me more creative in my approach and save me time?
Anticipated limitations
– I must take care to anonymise my data before sharing it with AI
– I may have more data than the maximum character count for uploads
– The AI may lack knowledge of the topic or individuals in focus
– Many tools may require subscriptions
The context
Research for a product, business and goal with which I am familiar but not yet an expert, especially at the start of a new project.
I brainstormed for 5-minutes to come up with a plan about where I could see value in using AI. I then consulted ChatGPT on tools to explore for each step. These tools I subsequently added to my original plan outlined below.
1. Desk research
Where AI can help
What can AI tell me about in-house knowledge that I might not know?
What can AI tell me about user priorities, interests, tastes and demographics that I’m not aware of?
How I’d go about this
- Inform AI that it is a senior user research, providing the context for the enquiry, and the specific goals for the study.
- Provide the AI with notes, transcripts, persona details or other materials from the business repository.
- Request the AI to:
- Analyse the data given the context and goal, summarising any key points to keep in mind during the research sessions.
- Highlight aspects that appear to contradict each other.
- List features or values of importance to the user group.
- Present information that appears vague or unfounded.
- Ask me what my assumptions are to challenge these assumptions.
I hope this could introduce new angles and help me adjust my expectations. I will
run through these ideas by the team or client.
Tool suggestions*
Consider using AI tools like Research Rabbit or Feedly for market research, alongside ChatGPT for summarising findings.
*I subsequently researched tools using ChatGPT’s suggestions.
2. Script creation
Where AI can help
As an experienced product professional I often have a useful script or template that has been tried and tested. I might also have a draft in mind, with questions noted down from kick-off meetings that I want to explore. How can I improve this script, given the project goals? What alternative approaches could I consider?
How I’d go about it
- Input an example script, and request this be used as a template, taking into account any further feedback I have on the script.
- Specify what’s needed for this project, new or different, such as:
- Length of questions
- Research goals
- Session duration
- Location
- Session format
- Types of question (open-ended and/or closed).
- If I have an existing body of work, I will anonymise it and ask AI to interrogate it for interesting angles to ask users about given the goals of the upcoming user sessions.
- If materials as conversation starters are useful, I might generate images as prompts.
- Review the suggested script and polish.
Tool suggestions
Tools such as Otter.ai for transcriptions and Google Docs for collaboration can enhance script preparation.
3. Session notes
Where AI can help
It is not easy to take quality notes and run the research session. Typing on a laptop can be distracting for the participant and hand-written notes can be hard to follow later. Sometimes a collegue is available for notetaking.
How I’d go about this
AI can provide a transcript and summary, which can be formatted into a table with the headers and rows you define.
I will also ask for a table to be generated from the script for each user, where I can write or type during the interview. This allows me to mark which questions and users provided interesting insights to highlight or explore further during the analysis.
Tool suggestions
Apps like Notion or Miro can help organise notes and insights visually.
4. Post-interview
Where AI can help
I’ve previously found that eight or more users back-to-back can be somewhat saturating. Without a strict note-taking or debriefing protocol between sessions, I often struggle to write up research later.
The AI-generated transcripts and summaries are the quickest method to capture a snapshop of what occurred. Knowing I have these tools to refer to post-interview is invaluable.
How I’d go about this
- Take a meeting transcript, including the user, date and time reference of the conversation.
- Request that it highlights any sentences that do not make sense so I can correct any unclear words due to sound quality, background noise, incomplete phrases, regional expressions or accents.
- Use this material to contrast against my own notes and shorthand.
- Interrogate the transcript for quotes, highlights, and common complaints.
- Share highlights with colleagues as a quick post-session update.
- Ask for possible patterns and conclusions, and then review them.
Tool suggestions
Descript is an excellent tool for audio/video transcription and editing that can be useful.
5. Analysis
Where AI can help
Analysis is arguably the most important part of all user research, however it is also time-consuming and there is great pressure to jump on the results. There is plenty of room here for efficiency-saving.
How I’d go about this
Using AI as a research partner, ask for different perspectives on the information to reflect and review:
- What words and terms are users using? I can incorporate these into the product or support materials
- What topics are new?
- What has changed in our assumptions?
- What opportunities does this new research present, if any?
- What could have been done better if the results were not so useful?
- Were there any surprises?
- Do these users match our existing profiles? In what ways? What is different about them?
- What should be the top priorities?
Tool suggestions
Text analysis tools like NVivo or Atlas.ti can assist in coding qualitative data, while ChatGPT can help synthesise findings.
What next?
This exercise led me to wonder, what if I ask an AI how it can help me with user research?
So I did:
If I have a user research project where I plan to interview 20 people for deeper insights into their use of a digital product, how can AI help me prepare for that?
You can find ChatGPT’s reply below.
Reflections
- Prompting AI makes me aware of how clear my instructions need to be. This has made me reflect on the way I communicate with colleagues and peers in general.
- AI can help you think through your work, whether it’s something you’re new to, or you’d like to brainstorm. It is a fabulous tool. It can also show you approaches other people have taken. This can be helpful, save time and be more creative, but it might not be useful for all projects and all occasions.
- The information you can share with it and the precision with which you ask for input are key success factors. See the results below, as well as the slightly uncomfortable cover photo for this post!
- Privacy questions, stacking and mixing different tools to reach reliable, coherent output, as well as personal approaches and budgets can make the results vary from one day and one person to another. It feels a little messy right now to be able to settle on one single toolset for my team. Experimentation and results sharing are inspiring and ongoing.
- Running experiments like this one for real-life needs is helpful in leveraging AI and learning fast.
- “One-shot prompting” is rarely enough. It’s usually necessary to reform a request to get to the information steps I have in mind. Having a conversation with AI is more fruitful.
ChatGPT’s take on my request
ChatGPT’s initial response to my question If I have a user research project where I plan to interview 20 people for deeper insights into their use of a digital product, how can AI help me prepare for that?
AI can be a valuable tool in multiple stages of your user research project. Here’s how it can assist you in preparing for interviews and optimizing the process:
1. Pre-interview: Understanding User Needs and Defining Research Goals
- Identifying Key Topics and Trends: AI tools can help analyze any existing data (e.g., surveys, user reviews, analytics) to identify trends or pain points in how users interact with the product. This will help you pinpoint the areas to focus on during your interviews.
- User Segmentation: AI can assist in segmenting users based on behavior, usage patterns, or demographics, so you can ensure that you’re interviewing a diverse sample with varying experiences.
- Creating Interview Guides: Based on the insights gathered, AI can help generate interview scripts or question frameworks that are tailored to uncover deeper insights related to user behaviors, motivations, and challenges.
2. During the Interview: Efficient Data Collection and Analysis
- Real-time Transcription and Analysis: AI-powered transcription tools (e.g., Otter, Descript) can transcribe your interviews in real-time, making it easier to focus on the conversation instead of manually taking notes.
- Automated Sentiment Analysis: AI tools can quickly analyze sentiment and tone during interviews, flagging areas of interest or emotional responses that may be significant, even if they are not immediately apparent to you.
- Recording Key Insights: AI can help organize and tag key insights, themes, or important user quotes during the interview, saving you time and ensuring nothing is missed.
3. Post-interview: Synthesizing and Analyzing Data
- Identifying Patterns: After the interviews, AI can help identify patterns and correlations between users’ behaviors and feedback, making it easier to group responses and extract meaningful themes.
- Creating Personas: AI can assist in synthesizing interview data into user personas that reflect the different segments or archetypes you’re investigating. These personas can then inform product design or decision-making.
- Automated Reporting: AI can assist in drafting comprehensive reports based on interview transcriptions, summarizing key findings, trends, and recommendations in a structured format.
4. Improving Efficiency & Reducing Bias
- Objective Data Analysis: AI can support the analysis process by reducing human bias. It can flag inconsistencies or biases in how you interpret responses, ensuring that the analysis is as objective as possible.
- Optimizing Interview Scheduling: AI-based tools can help automate scheduling and communication, optimizing the process of setting up interviews and managing logistics.
Would you like specific tool recommendations to help with any of these tasks, or more details on a particular phase?
When I asked for the tools for small teams or freelancers to avoid the unaffordable, it provided me with a useful selection, including pricing. However, naturally I should have specified a budget.
My learning journey continues!
Summary for Small Businesses & Freelancers
- For Text Analysis:
- MonkeyLearn and TextRazor offer free plans and are highly accessible for small teams.
- VaderSentiment is great if you’re comfortable with Python and want a completely free sentiment analysis tool.
- For Visualization:
- Power BI has a strong free version and is ideal for visualizing trends and making reports.
- Tableau Public is a good choice for freelancers who don’t mind public data.
- For Summarizing Large Text Data:
- SummarizeBot is a budget-friendly option for summarizing long interview transcripts or surveys.
- For Survey Tools:
- Qualtrics might be out of reach for very small businesses, but for a growing small business, their starter plans are a good option.
- For Transcription:
- Otter.ai is affordable and effective for freelancers who need quick and accurate transcriptions.
Key Takeaways:
- Best for Small Scale: Use MonkeyLearn, TextRazor, and Otter.ai (especially if you only need basic capabilities or are just starting).
- Best for Visualization: Power BI is the most accessible for freelancers, while Tableau offers better visualization features but with public data.
- Best Budget Option: For sentiment analysis, VaderSentiment is free, but if you want something more automated, MonkeyLearn‘s free plan is great for small businesses.
