What GenAI assistants and AI-powered developer tools have been most successfully used in your software engineering organization? Why?

In BASF we have applied GitHub Copilot as AI-support tool for software developers. Our user community is growing steadily, with mostly very positive feedback. The portfolio and performance of features is impoving over time, we have an established dialogue with MS to provide our input. The use of GitHub Copilot also helps to standardize the coding process within our company to a certain extent. The white spot we still have is ABAP code development.

We have used a variety of tools. We have standard deployment options of Cursor and GitHub Copilot that are the most used by engineers due to their integration and ease of use within existing deployment, CI/CD, and DevSecOps cycles. We also have an enterprise agreement for ChatGPT Enterprise which has Codex and use that in workshops to accelerate UX feedback cycles for base requirements gathering. 

We have embedded watsonx.ai to assist users articulate their functional area expertise which our experience revealed was the biggest impediment in capturing and matching human expertise. We had a great method to capture functional area expertise with fine granularity 100x to 1000x finer than any existing program, including LinkedIn. However, users, including highly qualified and experienced work seekers, needed Help Desk support in providing accurate and comprehensive inputs. Now with watsonx Gen AI Assistant, we are finding users making accurate and comprehensive inputs. This makes our most powerful solution cost-effective too. It is called PEMS. More details here: https://www.expresscomputer.in/guest-blogs/systems-dynamics-leverages-ibm-watsonx-ai-for-smarter-talent-acquisition-deployment/126453/
  

GitHub Copilot — best overall adoption + measurable velocity gains (e.g., controlled studies show ~55% faster task completion; internal enterprise studies also report faster merge times and higher dev confidence). Tight PR/issue integrations keep it in flow

I am having extremely good results with GitHub Copilot in VSCode agentic mode with Claude Sionnet 4. The observability and tooling allows for features to be created with good quality and the instructions file in vscode allows for a great level of control on style and practices.