Most nonprofit AI pilots I see are not badly chosen. They are badly handed over.
The demo goes well because the person running it built it. They know which prompts work, which files to point it at, and what to do when it returns something odd. None of that is written down, because it lived in their head the whole time it was being built.
Then the pilot ends. The tool passes to whoever has capacity. Within three weeks it is either abandoned or quietly producing work nobody checks.
What the handover actually needs
Not training. A written record of the decisions.
Why this tool and not the other one. What it is allowed to touch. What a bad output looks like, with an example. Who to ask when it breaks.
The person who ran the pilot knows all four. The person inheriting it knows none, and none of them are in the tool.
The test I use
Before a pilot ends, the person who built it takes a week off and someone else runs it. Not a handover meeting. An actual week.
Whatever breaks in that week is what your documentation is missing. It is uncomfortable and it is the cheapest thing you will ever do.
