NorthCode logo

The AI Colleague – Our Approach to the AI Revolution

NorthCode's AI Colleague uses company data to answer questions, optimize workflows, and enhance systems, paving the way for smarter, self-learning tools in business.
Photo Credit: Ismo Aro

The AI Colleague – Our Approach to the AI Revolution

AI is a rising tide that is increasingly affecting all areas of business. Even though the technology is still gathering momentum, it’s wise to prepare for major changes now and pragmatically harness the benefits of this simultaneously exciting and somewhat intimidating innovation.

At NorthCode, we call our approach the AI Colleague. This colleague has been trained to leverage your company’s contextual data from a variety of sources—for example, version control systems, wikis, CI/CD environments, chats, emails, and system logs. Modern large language models excel precisely because they can learn from vast amounts of data and communicate using natural, human-like language.

What exactly could an AI Colleague be in practice? At the core of the AI Colleague is a suite of applications built on top of this trained data. The simplest “low-hanging fruit” is a basic chat interface. Different stakeholders within the organization can pose various questions, for example:

  • A junior developer might ask fundamental questions about project practices.

  • A senior developer might need information about a legacy project that requires fixing, even if the original developers are no longer in the company.

  • Managers can inquire about project status and investigate reasons for potential delays.

Beyond the chat, the next step is more autonomous AI agents capable of doing more. They could, for example:

  • Provide code review feedback on pull requests.

  • Increase unit or acceptance test coverage.

  • Refactor code based on different parameters (clarity, efficiency, eco-friendliness).

  • Optimize CI/CD pipelines.

  • Suggest overall system optimizations.

  • Develop new features and even create pull request proposals.

Taking it one step further, these agents could potentially teach themselves. An agent could run experiments on the code, execute them in a CI/CD environment, analyze the results, and learn from them—producing continuous improvement.

While some of this may still seem like science fiction in Q1/2025 (when this was written), how much longer will that be the case? From our perspective, this should be taken seriously now—especially when designing and implementing Internal Development Platform (IDP) capabilities. Time will tell where this technological path leads us, but deep-learning-based chat solutions are already here today and are delivering value to us all.