video1.0<iframe src="https://www.loom.com/embed/c57b35c2580743d8b973340e3fedd009" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/c57b35c2580743d8b973340e3fedd009-394fc09e8b31ccb5.gif1164.163Anton and Donna, software engineers at Atlassian, discuss their experiences with using AI for software development. Initially, Donna was sceptical about AI's capability to write code, as her early attempts resulted in poor-quality outputs. However, over the past few months, AI technology has significantly improved, enabling it to pre-generate complex code for pull requests and automate repetitive tasks. They highlight the use of a tool called Rovo Dev Agent, an internal Atlassian tool that allows engineers to automate code generation through detailed prompts. This tool has been particularly useful for handling large volumes of repetitive tasks under tight deadlines, contributing substantially to their codebase like an additional team member. The conversation includes tips for writing effective AI prompts, emphasising the importance of treating AI like a new engineer on the team who needs clear, detailed instructions. They advise providing specific examples and leveraging the large context window modern AI offers to enhance the quality of generated code. Despite its advancements, AI still has limitations, such as forgetting imports, which requires human oversight to achieve final accuracy. They stress the importance of aiming for 80% accuracy in AI-generated code, with engineers completing the remaining 20%. The speakers also note the benefits of incorporating AI into their workflow, such as overcoming initial inertia in starting new tasks and aiding in onboarding new team members. They conclude with three key takeaways: write detailed prompts, treat AI as a team member, and leverage AI to pre-generate substantial portions of code to save time and effort.