Software Development

From instructions to pull request—automatically

How we built an autonomous development workflow that plans, builds, and submits features for review with a single prompt.

The Context

We built this for ourselves first

Utah Website Pros is a web development agency we operate. Before offering AI-accelerated development to clients, we implemented the workflow on our own codebase.

This case study documents that implementation—what we built, how it works, and the results we're seeing. We believe in using the tools we sell.

The Challenge

  • Already using AI coding assistants, but still manually orchestrating the planning, implementation, and PR creation steps
  • Quick wins and polish items rarely made it to the top of the priority list
  • Wanted to maintain quality through review while increasing development velocity

The Workflow

How it
works

01

Provide instructions

Describe the feature, fix, or improvement needed. Natural language, no technical specification required.

02

AI plans and builds

The system analyzes the codebase, creates an implementation plan, writes the code, and runs tests.

03

Review and merge

A pull request appears with all changes. Review the code, provide feedback, or merge when ready.

In Practice

What this looks like

1

The instruction

"Review the website for accessibility best practices and suggest any changes that might be needed."

2

The result

  • Added skip navigation links for keyboard users
  • Improved semantic landmark structure for screen readers
  • Enhanced focus indicators for keyboard visibility
  • Hid decorative elements from screen readers
  • Fixed color contrast issues for visual accessibility
  • Pull request ready for review and merge

The Results

What we're
seeing

1

Prompt to PR

Features go from instruction to pull request with a single prompt

100%

Human Review

Every change goes through PR review before merging

Continuous

Delivery

Ship features as fast as they can be reviewed

The Insight

"The value isn't replacing developers—it's removing the friction between deciding what to build and having it ready for review."

Human oversight remains essential. The PR review step catches issues, ensures quality, and maintains control. AI handles the implementation work; humans make the decisions.

Want this for your business?

We can implement the same workflow for your codebase. Let's discuss whether it's the right fit.

Devclock Assistant

How can I help?

Ask me anything about Devclock's services, pricing, or how we work.