👋 Hey Platform Leaders! I'm curious to see how other Platform/SRE teams are using AI to support their daily workflow and the tangible benefits of these specific use cases. So far, it seems there has been observed resistance and skepticism in groups I've interacted with in the platform engineering space. Some areas we are going to be experimenting with involve Copilot and Cursor in the IDE to facilitate writing Terraform and automation scripts for our infra. I'll continue to check out other Slack groups and online forums to gather more info but thought I'd reach out here to get some additional thoughts/ideas on how other leaders are experimenting with AI. Also, if you have any guidance on how to communicate the benefits and approach teams showing resistance to AI, please share them with me. Much appreciated!

I've been interested in this topic as well. Beyond the code assistants, we've been testing LLMs to help with support workflows. To date, it's not yet been a noticeable impact on our support load, however. I can't say the reasons are clear as to why, though I have some suspicions: 1. Currently the cost to reach a human is low, so people naturally default to that. 2. Some hallucination issues with the LLM responses. 3. Accurate / up-to-date documentation, which is the backbone for the LLM responses, still lags.

We have started researching and experimenting with using LLMs and tools around them to do actual code migrations and automatic adoption of new platform features - rapid adoption is still one of our bottlenecks in the org. This is a very interesting area so far (e.g. with grit.io or aider) and we are constantly trying to use more and more of it - we are also trying to use Claude and ChatGPT to do some custom migration patterns. Specific examples are migrations to our latest micro frontend services, design system changes, etc. What I have found so far is that everyone is super interested once they start playing around with it, but getting someone to actually start experimenting with LLMs for such use cases is still a big hurdle (due to initial bad experiences with hallucination etc).

@Patryk Bak is exploring this area within IDP in BlueSoft.com

@ma_ga this is really interesting! I know that OpenRewrite and a friend of mine who started Moderne drafted some thoughts on this as well. I want to believe we can get to a place where AI can largely handle the majority of refactoring cases (at least the fairly simple ones), but every time I've tried using it for side projects, I'm skeptical.

I do lots of stuff, and I write about them! like this