Hi, it's Mocha,
If this newsletter has one sentence that refuses to leave, it is this: AI drafts, you decide. I have written it about donor letters and support bots and PAC minutes, and today I want to give it a whole issue, because it is the difference between AI that builds trust and AI that quietly spends it. I hold this line in my own products (the meal scanner in my app is helpful when it knows and honest when it does not), and I hold it here: this newsletter takes shape with Claude every week, and every word still passes through my hands before it reaches you.
The three lists. The practical version of "human in the loop" is knowing, in advance, which of three lists a task belongs to:
AI alone is fine. Reformatting, summarising for your own eyes, first drafts nobody else sees, converting notes, checking a document against a checklist. If a mistake here costs you a shrug, let the machine run.
AI drafts, human reviews. Anything that leaves the building: newsletters, proposals, customer replies, grant answers, meeting minutes. The AI does the typing; a named human does the deciding. Named matters, because "someone will check it" is how nothing gets checked.
Human only. Judgement calls, money decisions, hard conversations, apologies, signatures, and anything where the relationship IS the work. A donor call, a performance conversation, a condolence note. AI can help you prepare for these. It must never perform them.
Build the loop into the plumbing, not the memo. Good intentions lose to busy Tuesdays, so make the workflow enforce the list. In automation tools like Zapier or Make, that means a draft lands in a folder or a "pending approval" step instead of auto-sending. In your assistant, it means the standing instruction "give me a draft to edit, not a final to send." In a team, it means the rule that no AI-touched text goes out under a name that has not read it. None of this is fancy. All of it is the difference between issue 7's quiet layer and the horror story where the wrong email hits the wrong list at 2am.
💚 The part I care about most. The loop is not a leash on the technology; it is what makes the technology restful. When you know the machine cannot take an irreversible step without you, you stop half-watching it, and the time it saves becomes real. And your people feel the same thing from the other side: customers, donors and parents can tell when a human stands behind the words. In a year when everyone's inbox is filling with obviously automated everything, "a person decided this" is becoming the scarcest signal there is. Guard it. It is your brand now.
If you try one thing this week (under 30 minutes): Write your own three lists for your role: five tasks where AI can run alone, five where it drafts and you decide, five that stay fully human. Stick it somewhere visible. You have just written your organisation's first AI policy, and it is a better one than most.
Your turn: Where do you draw your human-only line, and has anything ever tempted you to move it? I read every reply.
Mocha
AI enablement for mission-driven teams · Mocha's MindLab Inc.
Reply anytime. I read every one.
www.mochashmigelsky.com