Using AI to triage deal flow without handing it your judgment
AI is good at noticing that a message looks like a deal. It should not be the one deciding. A practical way to split the work.
What AI is good at here
Reading hundreds of messages and noticing which ones mention a company, a price, a raise or a request to talk. That is tedious for a person and quick for a model.
What it should not do
- Decide what is a deal. A model cannot know that the founder is a friend, or that you passed last year for a reason.
- Act on your behalf. Filing, replying and sending documents should stay with you.
- Hide its source. A suggestion without the message it came from cannot be checked.
Three rules that keep it useful
Review before anything is filed
Treat every suggestion as a draft. Nothing reaches your pipeline until you accept it.
Always keep the source
Every suggestion should link to the conversation it came from, so you can judge it in context.
Know whose key it is
Know which AI provider processes your messages and under whose account. Using your own key keeps that relationship yours.
How Merger does it
Claude features in Merger use your own Anthropic API key, billed by Anthropic separately from Merger. Claude surfaces potential deals with their source, and you review each one before it is filed on your Deal Desk or dismissed. You can use messaging and organize deals by hand without enabling Claude features at all.