A large telecommunications company wanted to pilot AI-assisted coding and how it fits their development workflows, and what benefits the technology could bring the company.
The challenge: Helping adopt a new technology paradigm
As part of its digitalization strategy, the client company wanted to evaluate what kind of value an AI-powered pair programming tool, namely GitHub Copilot, would bring to its business and on what scale. GitHub Copilot is a software tool that integrates with a developer’s code editor. It supports their daily activities by auto-completing code and providing learning possibilities by suggesting solutions based on code comments or direct questions.
Like most companies, our client wanted to accelerate the time it takes to introduce new digital services and features to customers while keeping the quality standards high. In addition, one of the critical requirements emphasized by the project sponsor was an improved developer experience. The ability to code and learn on the job without breaking the flow is vital to a productive and meaningful work experience.
The internal project manager of the client company commented: “Some of the most important drivers in improving our capability to provide valuable digital services are the developers’ satisfaction and flow state in their work. We want to learn how to improve developer experience with AI-powered tools.”
Eficode’s specialists have several years of experience with AI tools in development. The client wanted to utilize that experience in using GitHub Copilot to deliver on the promise that developers spend less time on repetitive tasks and improve testing capabilities and documentation. The first proof-of-concept project aimed to get initial developer insights and gain their support for the plan to expand the usage of AI tools in a controlled way.
The solution: A streamlined proof-of-concept
Eficode and the client have a long history of partnering, and we continuously look for new ways to ensure the client stays on the cutting edge in DevOps and software development to ensure and strengthen their competitive advantage.
A fast and efficient PoC was conducted to test the technology in real-life conditions. This helped build a clear business case of how Copilot could create a better developer experience and efficiency and value for the business. During the project, developers working with different languages and tech stacks kept a diary to gather insights from actual usage scenarios. Eficode’s team of consultants worked closely with the client’s development team every step of the way to ensure seamless integration and deployment of Copilot into their everyday workflow.
To drive wider-scale adoption, the project defined key performance indicators for tracking software development productivity, quality, and efficiency. In addition, a survey was created to track developer experience. These measurements are being used to understand how Copilot performs in actual use over time.
An observation: Guidelines and guardrails are needed
The project was concluded with a comprehensive report of the results, including an executive summary for C-level executives, practical insights and ideas from the participants, and a roadmap with suggested next steps. Some of the suggested future key activities included the creation of guidelines and guardrails to ensure an AI-assisted programming culture will start on the right groove for the client.
Some of the more actively debated topics around AI-powered tools revolved around intellectual property rights, data protection, and privacy. Eficode had already created our AI-assisted tools policy with our IT, security, and legal teams, giving us a prime position to help companies like the client create this documentation.
Results: Improved productivity and more satisfied developers
All the GitHub Copilot PoC participants from Eficode and the client supported continuing the use of Copilot and expanding the use in a controlled way. The participants agreed that Copilot users must follow clear guidelines to ensure the tools bring value to their work and not become a liability.
Many of the improvements that surfaced from the discussions during the PoC were around good DevOps practices, such as improved guidance on coding styles, test coverage, and keeping automated infrastructure provisioning up to speed when the coding work accelerated. Speeding up code creation requires the supporting capabilities to stay up to par to accelerate overall value delivery.
One of the developers who participated in the PoC says:
“Using Copilot to develop Java-based code for our integration platform brought several benefits, especially when I was not concentrating on the technology too much, but just let it help me implement the solutions I had planned to do at a faster pace.”