The insight
The spreadsheet is not the problem; the lack of structure underneath it is. Everyone in the business already knows how to read and edit a grid. Throwing that away and asking them to learn a form-based tool is a large cost for a benefit they did not ask for.
So keep the grid as the primary surface, and put a real data model behind it — typed fields, relationships, validation — without making anyone learn a new interaction model to get them.
What is worth stealing
Editing where the value lives is the default, not a feature. The grid is the editing surface, so there is no view mode and edit mode, no round trip, and no second mental model for the same object. This is the WYSIWYG principle taken to its conclusion.
Keyboard traversal is finished work. Arrow keys, Tab, Enter to commit, Escape to cancel, selection that survives the save. It feels like a spreadsheet because every convention people already have works exactly as expected — and that is a large body of unglamorous detail, not a coat of paint.
Views are per-person and per-purpose. The same table is a grid, a calendar, a gallery or a kanban, each with its own fields, filters, sort and grouping. This is the honest answer to the reality that operations, sales and finance want genuinely different slices — rather than one compromise table nobody likes.
Import is the front door. The data already exists. A tool that cannot accept it will run alongside the spreadsheet forever, and then the spreadsheet is still the real record.
Where it does not transfer
Flexibility has a curation cost. A base with forty views needs an owner, and without one it becomes the same mess as the shared drive it replaced. If nobody is going to prune, fewer, fixed views serve better.
The grid also assumes homogeneous records with comparable fields. For a record with genuine depth — a case, a contract, a patient — a grid row is the wrong container, and a hub page is right.
And "anyone can restructure it" is a feature until the structure is load-bearing for a business process. At that point you want the constraints a purpose-built tool provides.
The extraction
Meet people in the interaction model they already have, then add the structure underneath.
Before designing a form to replace a spreadsheet, ask what the spreadsheet was letting people do that the form will not — usually see everything at once, edit anything in place, and rearrange without permission. Those are requirements, not bad habits.