The Evolution of Software Projects: 100x Faster, 100x Less Work

A Journey from Hours to Minutes

I’m Matías Molinas, the CTO of Project Copilot. Eight months ago, we embarked on a journey to validate the core proposition of Project Copilot: helping to write the backlog in agile software projects, align code implementation with specifications, and validate the final implementation. To achieve this, we developed our own prompt workflow and utilized GPT-3.5 Turbo for evaluation.

The Initial Experiment

We created a small fictitious project (Fit Project) to assess the impact of using Project Copilot. The results were astounding. We reduced the work time from 10 hours to just 1 hour by leveraging two tools: Project Copilot and GitHub Copilot.

Evolution of Project Copilot

Evolution of Project Copilot

Today, we’re working on the next version of Project Copilot, which includes the same powerful assistant with enhanced capabilities. This new version is designed to create detailed plans executed through a series of checkpoints. Each checkpoint is implemented using a prompt workflow that can leverage our custom code interpreter. This component can call predefined functions to generate content and integrate it into project management platforms. Initially, this new version focuses on project management operations, such as the generation and review of project content and its evolution. It integrates seamlessly with platforms like Jira, Asana, and Azure DevOps. In the near future, we aim to extend its capabilities to include on-demand integration and analysis functions.

Integrating Chat with GPT-4o

We’ve also integrated a chat feature that wraps around GPT-4o, providing the power of GPT-4o and seamless integration with our workflow engine. This engine can turn plans into reality through real interaction and validation logic.

To validate our approach, we used the same project goal as our original fictitious project (Fit Project). Check out this chat example: Chat Example using a simplified version of the assistant in the new version of Project Copilot: Project Copilot GPT.

Remarkable Efficiency Gains

We realized that the work time reduced from 1 hour to just 5 minutes. The task, which previously took a human 10 hours, now requires only 5 minutes with the new assistant. Moreover, it doesn’t require more than writing the following prompt:

I want to create a client-side HTML and JavaScript app to calculate and display the health status score by comparing heart points and daily calories burned (based on daily steps, using the Google Fit API) with a healthy goal. Can you help me with that? I want to use Google Sign-In. I would like to create a plan using agile methodology and implement each user story with your guidance. We should use epics, user stories, and sub-tasks, implement the code, and validate it against the user stories.

And simply answering “Yes” five times!

The Future of Software Development

This progress showcases an incredible transformation in software development over just eight months. The way we develop software today is entirely different from the past, thanks to innovations like Project Copilot.

Stay tuned as we continue to revolutionize agile software projects with our AI assistant.

Matías Molinas
CTO, Project Copilot