When a new process starts, the pull is to buy a platform for it on day one. Then the process has to fit the platform, before anyone knows what the process is. My rule runs the other way. Do it by hand first, then in a spreadsheet, then automate it.
By hand
Run it yourself for the first customers. You find the exceptions, the questions people ask, and the steps that turn out not to matter. At Sheprd, a rideshare service for kids, I drove kids too, on a Massachusetts school bus license, and worked directly with parents, schools and afterschool programs. Doing the work is the fastest way to learn what the process needs.
Then a spreadsheet
Once the steps repeat, track them in a sheet. The columns you end up with are the data the process needs. The status column you keep editing is the list of stages. The notes column is where the exceptions live. You're writing the requirements without calling them that.
Then automate it
When the sheet starts to break, with two people editing at once, formulas the team has stopped trusting and the same order typed in three places, you know enough to automate it or pick software for it. The spreadsheet is the spec. A tool chosen at this point fits the process, because the process is written down. The signs are in 5 signs you've outgrown your spreadsheets.
When to skip ahead
Billing, contracts and anything regulated deserve a proper system from the first version, because a mistake there costs money or trust before anyone catches it. Everything else can earn its software.