Week 31: Automating user journey analysis with AI

The Experiment

User journey analysis can be mind-numbing. You know the drill: click through a homepage to sign-up flow, test product exploration paths, hunt for contact details, check page load times, mobile responsiveness, core web vitals—all the basics that help you understand what's actually happening on a site.

I had this thought: what if I could automate some of this tedious work? Not just to save time, but to see if AI could spot things I might miss (or at least handle the boring stuff while I focus on the interesting insights).

I wanted to test three key user journeys:

• Homepage to sign-up flow

• Product exploration paths

• Finding contact details

Plus the usual technical audit stuff—page load times, mobile responsiveness, and core web vitals.

The Process

Here's how I tackled it:

1. Discovering Playwright MCP I needed something that could actually interact with web pages, not just analyze static content. Within Claude, I found Playwright as an MCP server (Model Context Protocol—basically an agent way of doing stuff).

2. The Setup Dance I spoke to Claude about setting it up, then worked through terminal commands, created folders, and installed various packages. I must confess I didn't understand all of this stuff, but Claude guided me through it step by step.

3. Creating the Journeys Once set up, I defined the user journeys I wanted to test and let the system loose on the website.

4. Debug and Iterate When things didn't work (because they never work perfectly the first time), I just fed all the error details back to Claude and got it sorted. The back-and-forth was surprisingly smooth.

The Outcome

This turned out better than expected. The results gave me:

• Step-by-step breakdown of each user journey

• Performance metrics including page load times and core web vitals

• Detailed findings on buttons, navigation, and functionality

• Error reporting for broken images, navigation issues, and UX problems

• Overall assessment of user experience and performance

What surprised me most was how quick it was once set up. And what makes it great is that now that it's configured, I can run this same analysis on any website anytime. The iterative nature is brilliant.

Key Takeaway

AI-powered website analysis isn't just about saving time (though it does that). It opens up possibilities for agents to handle repetitive tasks that humans find tedious, while potentially catching details we might miss. Plus, once you've got the setup working, you can deploy it repeatedly without the manual drudgery.

Pro Tips for Beginners:

1. Don't Panic About Setup: You don't need to understand every technical detail. Let the AI guide you through the installation process.

2. Embrace the Debug Process: When things break (and they will), just feed the error messages back to Claude. The iteration is part of the process.

3. Think Beyond One-Time Use: The real value comes from being able to repeat these audits quickly across multiple sites or timeframes.

What's Next?

• I'm planning to test this on different types of websites to see how it handles various layouts and technologies

• Curious about extending this to more complex user journeys and conversion funnels

• I'm already playing with ChatGPT agents, watch upcoming newsletters

Want to Try It Yourself?

• Start with Claude and ask about Playwright MCP setup

• Define simple user journeys first—don't try to automate everything at once

• Be patient with the setup process, but know that the payoff is worth it for repeated use