I've been managing in the DevOps space for about 5 years now and have been quite hesitant about adopting AI technologies. My reasons include concerns over inaccuracies and the potential for skills atrophy among junior staff, as I believe it's important for them (and myself) to continue developing our skills. Recently, my boss and the executives have been pushing hard for AI adoption, leading to some heated discussions where I felt pressured to either adapt or leave. So, I'm looking to approach this cautiously by incorporating AI into our software development life cycle (SDLC) in a manageable way rather than fully replacing traditional methods.
I've heard that successful AI implementation focuses on improving accuracy, performance, and optimizing token usage. Personally, I've found AI to be effective for repetitive tasks when provided with examples or models. I also learned that 'Skills' can be used to define those repetitive tasks.
1. Has anyone created a repository similar to a devops-toolkit that shares "Skills"? What are the downsides? Do they require extensive context?
2. Has anyone implemented AI-driven workflows in their tools? Was it beneficial?
3. Are teams fully shifting their IaC repositories to incorporate AI? How about adding spec documentation to modules or using AI agents in CI/CD pipelines?
I welcome any helpful examples or stories to guide me in implementing these technologies moderately.
1 Answer
At my company, we've also been pushing for more AI integrations this year. Here’s what we've achieved so far:
1. Automated AI reviews for pull requests targeting the dev branch.
2. Using AI to help create IaC through IDE plugins—though it’s not perfect, it gets us about 80% of the way there.
3. We implemented AI within our IDE to help identify security issues and manage outdated packages.
4. We're piloting AI for auto-remediation of security concerns through pull requests.
5. Another pilot involves AI generating documentation and architecture diagrams based on the application’s source code.
We're developing more initiatives slated for 2026, but the groundwork has been laid!
That’s similar to what I’m trying to do, but my team has blocked AI-generated pull requests for security reasons. I noticed that while AI can help with generating docs and diagrams, we still need human intervention for certain tasks, especially in IAC. Do you automate documentation uploads with your CI/CD processes?