AI Agent for qualitative research

An AI Agent to speed up qualitative user research analysis.

Context

What does it do ?

To keep in mind

The flow

Time saved

Next step

Tools

As part of a skydive jump log project with the French Skydive Federation, where I’m working as a UX/UI Designer, I built an AI agent using n8n to automate the analysis of user interviews, which is a time/money-consuming step.

In just a few seconds, it can identify patterns, problems, and generate actionable insights. Essentially, it is an affinity diagram partner. The same workflow can be done with Notion or Airtable, according where you store your user interviews.

Despite the time saved, there are still important steps that need to be done manually:

  1. Identify the data we really need and remove what’s unnecessary.

  2. Review the results to check for bias.

  3. We still need to go over the results to ensure they make sense, as AI can still make mistakes: "Treat AI's work as a first pass, double-check their facts, and provide specific instructions."

  4. The AI prompt is crucial. It’s what generates the result. Therefore, it’s necessary to go through several iterations to achieve the desired outcome.

  1. Trigger the workflow on each update

  2. Get the user interviews from the Google Sheet

  3. Merge all rows in one single content

  4. Let the AI make the report according the prompt

  5. Create the HTML Version of the report

  6. Convert the HTML into a PDF

  7. Upload the PDF into Google Drive for the team

The time required to prepare and analyze user interview data and create an affinity diagram depends on the number of interviews and data complexity. Here’s an average estimate:

1. Preparing the data (~1 to 3h)

  • Transcribing interviews (if not automated): 30 min - 1h per interview

  • Summarizing key insights: 1 to 2h depending on volume

2. Creating the Affinity Diagram (~2 to 6h)

  • Grouping data on sticky notes or a digital tool (Miro, FigJam): 1 to 3h

  • Identifying patterns and themes: 1 to 2h

  • Discussing and validating with the team: 1h

3. Total estimated time:

🕒 Between 4 and 10 hours, depending on the number of interviews and the depth of analysis.

With my AI Agent, it takes only minutes. In other words, I save between and 2 days of work.

The next workflow would be similar: a workflow to analyze quantitative research with data visualization for Google Forms responses.

n8n, HTML & CSS, PDFShift, a bit of JavaScript, and some APIs.