This could be a good approach. Let people figure out optimal workflow with their spreadsheets, and when those spreadsheets stabilize, then help them move to a database. Aside from data entry speed, part of the reason is flexibility, especially when you're just figuring out what's the best way to collect and process data for a particular task.
As a developer in a company that does custom integrations, I wholeheartedly agree. Let them pay us to convert their unstructured files, instead of letting some automated importer do the job.
I'm only half kidding; the margin on our Pentaho Kettle transformations is quite good.
