How J.P. Morgan developers leverage AI

AI is making waves across industries, and that's perhaps most evident in the field of software development. Given the fast-paced nature of software development, developers need access to the latest technologies and tools to maintain a competitive edge. At J.P. Morgan Payments, we've integrated AI-powered developer tools such as PR review tools, as well as LLMs to streamline workflows, enhance code quality and optimize our software development life cycle.

AI code assistants

AI-powered code assistants have transformed the way developers approach coding, making the process more efficient and intuitive. Tools like GitHub Copilot, Codeium, DevGPT and Tabnine directly integrate into development environments, offering real-time code suggestions and completions. By harnessing advanced machine learning models, these assistants can predict and generate code snippets that are contextually relevant to the project at hand. This capability not only can help accelerate development time, but can also help enhance code quality, allowing developers to concentrate on solving complex problems rather than getting bogged down by repetitive coding tasks and boilerplate code.

For instance, at J.P. Morgan Payments, we utilize JetBrains IntelliJ IDE, renowned for its robust features such as intelligent code completion. By integrating GitHub Copilot, we further enhance our development process, enabling our team to work more efficiently and creatively. This combination of tools empowers developers to push the boundaries of innovation, ultimately leading to more sophisticated and reliable software solutions. As AI-powered code assistants continue to evolve, they promise to redefine the landscape of software development, making it more accessible and dynamic than ever before.

AI code review and PR tools

In parallel with the advancements in code assistance, AI is also revolutionizing the code review process, an essential component of software development. Code reviews play a crucial role in maintaining high code quality by ensuring that multiple team members approve code before it is deployed to production. However, the traditional review process can be time-consuming, as it requires several people to read, review and test the code. This is where AI tools can help make a significant impact. AI-powered pull request (PR) and code review tools can help efficiently analyze code changes, provide insightful feedback and enforce coding standards. By incorporating AI into the review process, bugs and predefined issues can potentially be identified and addressed before a human reviewer even looks at the code.

At J.P. Morgan Payments, we have developed an internal tool and workflow called PRBuddy to enhance our code review process. PRBuddy leverages AI to streamline the review and management of pull requests by offering intelligent feedback and suggestions. It automatically generates detailed PR descriptions, including titles, types, summaries, code walkthroughs and labels. Furthermore, it identifies modified code components, facilitating documentation and offering code suggestions. By integrating PRBuddy into our workflows, we not only expedite the review process, but also help ensure a higher standard of code quality, ultimately leading to more robust and reliable software solutions. As AI continues to advance, tools like PRBuddy will become increasingly vital in optimizing software development practices.

Large language models (LLMs)

The recent surge in interest around artificial intelligence can largely be attributed to the widespread adoption of large language models (LLMs), particularly following the release of ChatGPT. When OpenAI launched ChatGPT in 2022, powered by their GPT-3 model, it brought the capabilities of LLMs into the mainstream, showcasing their potential to transform various industries. However, alongside the benefits, there are inherent risks associated with this technology, including data leakage, compliance issues and security vulnerabilities.

Rather than avoiding the technology due to these concerns, J.P. Morgan has chosen to embrace it responsibly by developing an internal LLM called LLM Suite. This approach allows us to harness the power of LLMs while mitigating the associated risks. LLM Suite is tailored for specific use cases, such as software development, where it aids in tasks like code review, writing unit tests and brainstorming, thereby helping to enhance the development process. An internal LLM such as LLM Suite can help you safely and effectively integrate AI into your operations. This can help drive innovation and help you effectively maintain a high standard of security and compliance. As AI technology continues to evolve, our commitment to responsible adoption ensures that we remain at the forefront of industry advancements.

Stats on AI impact on developer productivity

• Speed of task completion: A study by McKinsey indicates that software developers can complete tasks twice as fast when utilizing generative AI tools. This is due to efficiency gains in their overall workflows.
• Collaboration enhancement: According to a survey reported by GitHub, 81% of developers believe AI coding tools will improve collaboration within their teams. This includes enhancing processes like security reviews, planning and pair programming.
• Tool integration: Generative AI is recognized as a valuable addition to the developer toolkit. One of the main reasons it is poised to help developers in particular is its ability to be integrated into the tools developers already use, such as IDEs, according to IBM.

Conclusion

As newer, more advanced and more capable AI models are released, there will be newer use cases and efficiencies to be gained by developers. Integrating AI into the software development process at J.P. Morgan Payments has proven to be beneficial. We’ve not only helped streamline workflows, but also helped enhance the quality and reliability of our software solutions.