I see Excel as the weak point in Office, at least from the viewpoint of getting the right answers.

There was that time I went and gave a talk in NYC and then came back to give the same talk in Ithaca and had an eagle-eyed web developer note an error in a chart I made with Excel because of the broken defaults.

There are numerous ways people get wrong answers with Excel, plus Excel does some really strange stuff to hide the fact that

  0.1 + 0.2 != 0.3

in floating point numbers (hint: none of those numbers actually exist in the floating point number system, when you ask for 0.1 you get a different number that happens to print out as '0.1')

There is something brilliant in the spreadsheet model, namely, it frees programming from the need to sequence operations in some particular order. The CS community has seen relaxing this as a way to find parallelism for the CPU & friends but not so much as a way to take a burden off the programmer who might not need to think about it.

However there is quite a bit structurally inadequate about Excel from the grid model which doesn't necessarily fit the problem and no support for software eng practices to make complex sheets that really work.