> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getoneprofile.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Reply with AI

> Have a draft written for you, check it against the thread, and approve it before anything sends.

Reply with AI writes a draft reply from the thread, the sequence's goal and your business context. You read it, edit it if needed, and decide whether it sends.

<Frame caption="A drafted reply waiting for approval in the inbox.">
  <img src="https://mintlify.s3.us-west-1.amazonaws.com/oneprofile/images/outreach/reply-with-ai.png" alt="An AI-drafted reply card above a message thread, showing a confidence score, an editable draft body, and Approve and send, Regenerate and Discard buttons." />
</Frame>

## Generate and review a draft

<Steps>
  <Step title="Click Draft a reply">
    Oneprofile reads the conversation, the sequence's goal and your business context, then writes a draft. Nothing sends yet.
  </Step>

  <Step title="Read the confidence score">
    A high score means the draft is ready to send as is. A lower one is worth a closer read first.
  </Step>

  <Step title="Edit if needed">
    The draft is a plain text field. Change anything before you approve it, or click **Regenerate** and describe what to change instead.
  </Step>
</Steps>

## Approve, skip or discard

**Approve and send** sends the draft immediately with any edits you made. **Skip for now** keeps the draft and brings it back next time you open the thread. **Discard** deletes it, and rating a draft helps future ones.

<Warning>
  A sequence set to auto-pilot can send a high-confidence draft on its own, without anyone clicking approve. Anything the model is unsure about still waits for a human.
</Warning>

## Set the goal and tone

A sequence's **AI replies** section, on its Settings tab, sets what each draft is trying to do. It offers a **Goal for this conversation**, **Tone**, **Language** and a **Call-to-action link**. **Reply mode** chooses between writing as a human, holding every draft for approval, or auto-pilot.

## Good to know

<AccordionGroup>
  <Accordion title="Can AI ever send a reply without approval?">
    Only on auto-pilot, and only for a high-confidence draft on a healthy thread. After a handful of auto-sends on the same thread, it goes back to waiting for a human.
  </Accordion>

  <Accordion title="What happens if I edit a draft and then regenerate it?">
    Oneprofile asks first, because regenerating writes over the draft on screen and your edits cannot be recovered.
  </Accordion>

  <Accordion title="Where do goal and tone come from if a thread has no sequence?">
    Reply mode, language and tone fall back to a workspace default. Goal and the call-to-action link do not, so a thread outside any sequence gets a plainer draft.
  </Accordion>

  <Accordion title="Can I save a draft as a template?">
    Yes. **Save as template** keeps the wording for reuse on future replies.
  </Accordion>
</AccordionGroup>

## Next steps

<CardGroup cols={2}>
  <Card title="Inbox" icon="inbox" href="/outreach/inbox">
    Read and reply to every conversation a sequence starts.
  </Card>

  <Card title="Business context" icon="sparkles" href="/settings/business-context">
    Tell Oneprofile about your company so drafts write like you.
  </Card>
</CardGroup>