>My personal hope is financial analysts get more tech savvy over time.
Probably won't happen and isn't important imo. If somebody can dig SQL/Python/R then somebody can get paid more to do something other than count beans (speaking here of lowest level financial analysts [accountants]; I was probably the only person at the division trying to make a meaningful financial forecast that was more detailed than a top level number with zero transparency. I started w/ python but wanted a COTS solution. Didn't want to babysit it.).
>Better use of databases would allow one spreadsheet with a database to store the changing inputs.
This is the key. The number one complaint of every other business area was that forecasts came in asynchronously and it was impossible to tell whether what had been added or removed was real. Hallway conversations took the place of structured data. That's fine for many things, but not for projections which determine whether we need to hire twenty more engineers or let go of a dozen.
The product that would've solved the problem (and then some) was IBM's TM1 (now "Planning Analytics powered by TM1"). My model worked well in TM1 and could've been extended to being better than a calculated model--the direct inputs from the dozens of IPTs could have been collected without the need for any math used to back out of numbers. No "model", merely the collection of demands of every program and every potential contract. The program teams already generated all of this data; finance threw most of it away while rolling it up.
TM1/PA has Excel plugins which render the change invisible to the financial analysts doing the work. Their job doesn't change but the organization can hoover up more data by at least an order of magnitude, breaking the cycle of relying upon opaque, top-level numbers that no other business area actually trusts.