DM handling is configured per agent under Agents → [your agent] → Settings → DMs. The DMs tab appears only for teams enrolled in the inbox beta.
The page has two sections: DM Handling (toggles) and DM-Specific Instructions (a textarea).
The Enable DM handling toggle controls whether the agent prepares AI drafts for inbound DMs on the pages it's assigned to.
Turning it off is also a clean way to silence the agent on a specific page without disabling the page entirely.
When this toggle is on, the first AI-drafted reply of each conversation includes a brief disclosure that an AI helped draft the response. Something like "An AI assistant helped draft this so we could get back to you fast." Subsequent replies in the same conversation do not repeat the disclosure.
Some industries (healthcare, finance, anything customer-trust sensitive) appreciate this transparency. Most consumer brands leave it off.
The textarea under DM-Specific Instructions (up to 1500 characters) is layered on top of the agent's general brand context, but only when drafting DMs. Use it for things that apply specifically to private conversations, where the tone, content, or rules differ from public comment replies.
Concrete examples that work well here:
If the user asks for our phone number, ask them for theirs and let them know a team member will call within 24 hours.
When sending a checkout link, always remind users that 10% off their first order applies automatically at checkout.
Never share order status. Direct the user to log in to their account dashboard.
These take priority over conflicting guidance in the agent's general additional info. The agent reads them as DM-only rules.
Settings save when you click Save at the top of the agent settings page. New drafts pick up the changes immediately. Drafts already in the inbox keep the text they had at the time they were generated. If you want to regenerate them with the new instructions, click Regenerate on the draft.
NEXT ARTICLE