I've been in the first type of work my whole academic career. I'm definitely more of an informatician/sysadmin than a scientist, and despite being entrenched in the science for almost a decade I can tell you that I never developed a 'feel' for it - it was too specialised, and I lacked the ground knowledge to bridge the gap. A smarter man would have found the time, but in my defense even in a single field has too many subfields to delve into, and you simply cannot master them all.

The best projects, and when I'm most efficient at work, is when someone explains the input and describes the desired output in as most abstract way possible.