How I work
Getting AI past the demo
A few things I've learned about getting AI from an idea to something people use.
1.
Start from the number someone already cares about.
Usually revenue or time saved. If the work can't move it, it probably shouldn't be built yet.
2.
Look at what's there first.
Most of the real constraints are already sitting in the data and in how people work.
3.
Build the smallest version that can meet real users.
A prototype in someone's hands teaches more than a polished plan.
4.
Stay for the first weeks of use.
That's where most of the useful fixes come from.
5.
Leave it with the people who'll run it.
Write it down and train whoever owns it next, so it keeps working after I've moved on.
If you're working on something like this, I'm always glad to compare notes.