08-11-2025
How to automate manual tasks with Windsurf Workflows
In the age of AI tools, it’s been delightful to find ways for AI to help boost my productivity. Whether it’s ChatGPT, Claude,
or an agentic IDE, it’s crazy how much they have augmented my development workflow. For example, as soon as my company approved Windsurf
as an IDE, I went ahead and experimented with it. My experience using Cascade (which is the name of Windsurf’s chat interface) and Windsurf’s autocompleting
feature have been super helpful. My main use of Cascade has been treating it as a debugging, coding partner in which I can
ask it questions related to the code base or ask it to implement a small, tedious task.
The magic though, is that second use case I highlighted: ask it to implement a small, tedious task.
And so, this brings me to an example of how I leveraged Windsurf Workflows at work to automate editing a large set of static files with repeated entries.
At work, we have microservices spread across multiple AWS regions that contain YAML files. There are instances where we need to edit those files, however the number of edits
are multiplied by the number of AWS regions we service. So, if we want to edit a microservice but there are 10 regions, that means manually editing 10 files.
That’s a lot of boring, manual work!
Noticing that this manual task could be automated, I went ahead and configured a Windsurf Workflow for my team. For this workflow to be successful, I ensured that my prompt was
specific, concise, and that there were clear steps that the LLM could do. I even instructed for it to pause and wait for acknowledgement from the prompter before it could continue doing the next step.
Once it’s done editing the files, I will then verify the changes and spin up a PR with those changes.
I was amazed and thrilled by the results, and my team and I have continued using this ever since. This has unlocked and freed up time for me to focus on more energizing tasks. It’s so nice to have
one instance of Windsurf doing these manual edits off to the side while I have a separate instance that I use to continue coding.