Why I wrote it
I wanted a fundraising team of agents. I already had a few kinds: a set of advisor bots I use like a board of directors for judgment calls, some always-on agents for scheduled work, and a coding agent.
My first two messages were short: I want a fundraising squad, and include a CRM agent. The third one changed the result. I stopped choosing a platform and asked for the best fit for each job.
What makes it work
It asks for a side-by-side placement. Without it, everything lands on whichever tool you mention first.
It asks for a written plan for sign-off, so the decision stays yours.
It asks the AI to flag anything already broken that the plan depends on. In my case that turned up a sign-in problem with the advisor bots before anything was built on top of them.
The prompt
I want a [fundraising] squad. I have [tool A], [tool B] and [tool C]. For each job the squad does, propose which one fits best and why. Write it as a plan for my sign-off. Nothing gets built or installed until I approve it. Flag anything already broken that the plan depends on.
How to use it
List every tool you have, even the ones you don't think fit. Describe each in a few words if the AI might not know it, for example "always-on agent that runs on a schedule" or "advisor bots for second opinions".
When the plan comes back, look for jobs with no good fit. That's often where you need a new tool, or where a person should do the job.