VibeOps

I Built a Professional CV Using Only an AI Browser Agent — Here’s What Happened

How agentic AI transformed a Gmail attachment into a polished Google Docs resume by navigating multiple tabs, extracting LinkedIn data, and formatting everything autonomously.

The Challenge

I had a CV template as an image attachment in Gmail. I wanted to recreate it in Google Docs with some modifications:

  • Replace the photo placeholder with my contacts
  • Replace the green header bar with my name
  • Pull my actual skills from LinkedIn
  • Format work experience with bold company names and italic dates

Simple enough for a human, right? But I decided to let an agentic browser AI handle it entirely.

What is an Agentic Browser?

Unlike traditional chatbots that only generate text, an agentic browser AI can:

  • See web pages through screenshots and DOM parsing
  • Navigate by clicking, scrolling, and typing
  • Work across multiple tabs simultaneously
  • Execute complex multi-step workflows autonomously

Think of it as giving an AI intern access to your browser — except this intern never gets tired, never loses focus, and can process visual layouts instantly.

From Image to Document:Analyzing the Template

The AI opened the Gmail attachment and captured a screenshot to understand the layout: a simple two columns one with photo, title, skills list and the other name, description, experience sections

Creating the Document Structure

In Google Docs, the AI:

  1. Inserted a 2×1 table
  2. Set the left column width to 5cm
  3. Applied a custom teal color (#2D5A5A) to the background
  4. Removed table borders for a clean look
textRight-click → Table options → Cell background color → Custom → #2D5A5A

Cross-Tab Data Extraction

Here’s where it gets interesting. The AI opened my LinkedIn profile in a new tab and navigated to the Skills section. It scrolled through 74 skills, identifying the most relevant ones:

  • Docker, Kubernetes, Azure
  • Git, Python, Jenkins
  • Linux/Unix, OpenShift, PostgreSQL
  • Scrum/Kanban, LangChain/RAG, Kafka

Then it switched back to Google Docs and typed them into the Skills & Tools section.

Formatting Work Experience

The AI processed my work history:

  • Bold for company names
  • Italic for employment dates
  • Removed verbose tech stack listings per my request

It used keyboard shortcuts (Ctrl+B, Ctrl+I) and precise text selection to apply formatting.

Making Links Clickable

Several certification URLs were plain text. The AI selected each one and pressed Ctrl+K to convert them to hyperlinks — a tedious task that took seconds instead of minutes.

The Struggles

It wasn’t flawless:

  1. Table height misconfiguration: Setting row height to 25cm broke the layout. The AI had to Ctrl+Z and retry.
  2. Text selection challenges: Google Docs’ text selection can be finicky. Triple-clicks sometimes grabbed the wrong paragraph.
  3. Scrolling through LinkedIn: 74 skills required multiple scroll actions. A smarter approach would have been bulk text extraction.
  4. Color picker navigation: The nested menus in Google Docs required several clicks and scrolls within the sidebar.

But here’s the key insight: the AI recovered from every mistake autonomously. It detected when something went wrong, undid the action, and tried a different approach.

Why This Matters

1. Complex Tasks Become Conversational:Instead of learning Google Docs’ table formatting quirks, I just said: “Make the left column dark green with white text.”

2. Cross-Application Workflows Are Seamless:Extracting LinkedIn skills → pasting into Google Docs → formatting text. Three apps, one conversation.

3. Tedious Tasks Disappear:Converting 6 URLs to clickable links? Formatting 4 date ranges in italic? The AI handled it while I watched.

4. Visual Understanding Changes Everything: The AI didn’t just read the template description — it saw the two-column layout, understood the color scheme, and replicated the visual hierarchy.

The Limitations

Let’s be honest about what doesn’t work yet:

  • Speed: Each action takes a few seconds. A human expert might be faster for simple tasks.
  • Precision: Text selection in rich editors sometimes requires multiple attempts.
  • Visual-heavy apps: Tools like Figma or Canva are harder to navigate than text-based interfaces.
  • Authentication: The AI can’t log into new accounts or handle CAPTCHAs.

The Future of Work

What I experienced is a glimpse of how knowledge work will evolve:

  1. Describe outcomes, not steps: “Create a CV matching this template” instead of “Insert table, set column width, apply color…”
  2. AI as a browser co-pilot: Watching an AI navigate tabs, click buttons, and fill forms feels like having a tireless assistant.
  3. Compound tasks become trivial: LinkedIn → Google Docs → Gmail? One conversation.

Leave a comment