Design decision I keep coming back to: AI drafts the reply, a human approves before send
Summary
The article discusses a design decision where AI drafts a reply for users to approve before sending, emphasizing a human-in-the-loop approach to maintain control and quality.
Similar Articles
Superhuman’s new auto-draft feature almost makes me like AI replies
Superhuman launches a new auto-draft feature that uses frontier models from Anthropic and OpenAI to generate email replies that match the user's tone. Early beta testing shows 40% of auto-generated drafts were sent within a day, with 60% sent without manual editing.
Stop building autonomous email agents
The author argues against building fully autonomous email agents based on real-world failures, advocating for constrained 'propose-and-approve' workflows where AI prepares context and drafts but humans retain final approval on sends.
@imryven: I clicked "approve" on my own AI ninety-three times out of a hundred without reading a single word of what I was approv…
The author discusses how frequent AI prompts cause users to ignore them, and highlights a team that reduced prompts by 84% through system redesign to eliminate unnecessary options.
I built a human approval inbox for AI agents after writing the same glue code three times
Developer introduces Impri, an open-core human approval inbox for AI agents that prevents unauthorized actions by enforcing structural gates instead of prompt instructions.
“Human approved” might be hiding a much harder governance question
The author argues that merely proving a human clicked Approve in an AI system is not the same as proving the approval was legitimate — and that the real governance challenge is establishing why that human's decision was authoritative at a given time. It asks how builders of high-consequence agentic systems should think about eligibility, assignment, authority, independence, and timeliness.