Nobody loves writing documentation. It’s one of those tasks that always gets pushed to the end of a project, squeezed in right before handoff, and takes longer than it should.
But clients need documentation. They need to know how to update their homepage banner, add a new team member, or change their business hours. Without it, you become their documentation, fielding emails and calls for tasks they could easily handle themselves.
I’ve been using AI to streamline this. Here’s the workflow I’ve developed that turns documentation from a time-consuming chore into a 30-minute task.
The Problem with Traditional Documentation
Writing client documentation manually means:
- Logging into the WordPress admin and navigating through every screen
- Taking notes on what each section does
- Writing instructions that make sense to someone who doesn’t live in WordPress every day
- Remembering to update the docs when you customize anything
- Finding time to actually do all of this
That last bullet is where most of the friction lives.
The AI-Assisted Approach
My workflow uses Claude Code (Anthropic’s command-line AI tool) to generate documentation from screenshots and project context. It’s straightforward and produces surprisingly thorough results.
Step 1: Capture the Admin Interface
I take screenshots of every WordPress admin screen the client will interact with. This typically includes:
- The dashboard
- Posts/Pages editing screens
- Any custom post types
- Theme customizer or block editor settings
- Plugin interfaces they’ll need to use
- User management (if applicable)
Quick, full-screen captures are fine. The AI is good at interpreting what it sees.
Step 2: Feed Everything to Claude Code
With Claude Code running in my project directory, I provide the screenshots along with a prompt like:
Generate client-facing documentation in Markdown format for this WordPress site. Use what you know about the project structure and these admin screenshots. Write for a non-technical user who needs to manage their own content.
Because Claude Code has access to the project files, it already understands the theme structure, what plugins are active, any custom functionality I’ve built, and how everything connects. The screenshots fill in the visual context of what the admin interface actually looks like.
Step 3: Refine in Google Docs
The AI generates solid Markdown documentation, but it’s a first draft. I import it into Google Docs where I can:
- Adjust the language to match the client’s technical comfort level
- Add any client-specific notes or preferences
- Format it for easy reading with headers and spacing
- Share it directly with the client or export as a PDF
The heavy lifting is done. I’m editing and refining, not writing from scratch.
Why This Works
The AI brings two things to the table that make this effective:
First, it has full context of the project. When Claude Code is running in my project directory, it can reference the theme files, plugin configurations, and custom code. It knows this isn’t a generic WordPress site. It understands the specific implementation.
Second, it’s good at translating technical reality into plain language. The screenshots show what the interface looks like; the AI describes what it does in terms a client can understand.
Bonus: Developer Documentation
The same approach works for internal documentation. After generating the client docs, I’ll often follow up with:
Now generate internal developer documentation for this project. Include the theme structure, custom functionality, hooks and filters, and anything another developer would need to know to maintain this site.
This creates technical docs that explain the “why” behind implementation decisions, useful for future me or anyone else who inherits the project.
The Time Investment
The whole process takes less than 30 minutes:
- 5-10 minutes capturing screenshots
- 5-10 minutes prompting and generating
- 5-10 minutes refining and formatting
Compare that to 2-3 hours writing documentation from scratch for a typical site. That’s time I can put toward actual development work.
A Few Tips
After running this workflow on several projects, here’s what I’ve learned:
Be thorough with screenshots. It’s faster to capture a screen you don’t need than to regenerate documentation because you missed something.
Include context in your prompt. If the client has specific terminology they use, or if certain features are more important than others, mention it.
Don’t skip the refinement step. AI-generated content is a strong starting point, but it needs a human pass. You know your client; the AI doesn’t.
Save your prompts. Once you find wording that produces good results, reuse it. Consistency across projects makes your documentation more predictable and professional.
The Bigger Picture
This isn’t about replacing the documentation process. It’s about removing the friction. When documentation takes under 30 minutes instead of 3 hours, it’s easier to fit into every project.
Clients get the resources they need. You get fewer support requests. Everyone wins.
That’s the practical side of AI in client work: not flashy, just useful.