The task here is simple: to reduce work, it's a simple investment. The choice of picking the correct bottleneck is as important as the tool itself, as this gives it use. If it's never used, it never creates value. The same goes for uptime, it's important that stability is almost perfect, as downtime undermines the utility of having the tool. Remember that making and using a tool also creates debt for future maintenance of the tool.
To adhere to this, I often choose Golang or Python as they are cheap to maintain. Golang has good stability because it is easy to code and compile checks give early error detection. A word of caution on AI generated tools: while they are cheap to make, the maintenance debt grows exponentially, as the code is often poorly understood by the team and difficult to reason about over time.