Missive's AI assistant reads a thread's internal comments too

The setup

Missive is built around shared team inboxes where a thread already carries internal chat comments and assignment history alongside the customer-facing email, so an AI assistant drafting a reply there has more to see than a single-user mail client would. Missive's own documentation on the Assistant states it requires the Productive or Business plan, and only an organization admin can add or manage an AI provider, which is linked to the whole organization rather than one person's account. Setup runs through either Missive-managed AI credits or a provider API key a team supplies itself.

What the documents show

The documentation is specific about scope: when opened from a conversation, the Assistant sees every email, internal chat message, and note in that thread, which is the internal context the topic's premise anticipated. A companion overview article states the Assistant can also search across all connected email accounts, check calendars, and look up canned responses when asked, but it draws a line the thread-context claim does not cross on its own: reaching outside the open thread requires either attaching that data explicitly with an @ mention or asking for it in a prompt, not automatic mailbox-wide access by default.

The friction

Cost is split two ways and the documentation is blunt about the trade-off: Missive-managed credits carry a 25 percent markup over the provider's own price but need no API key, while bringing your own key means paying the provider directly with no markup but managing billing and limits outside Missive. Organizations on a qualifying plan start with five dollars of free credit, not an unlimited allowance, and when a credit balance hits zero, the documentation states AI stops working until an admin tops it up. Every provider connected also carries its own data-retention policy that the documentation directs an admin to review separately.

What changed in the work

The documented change is that drafting a reply in a shared inbox can now start from a thread's full internal record, not just the customer's own words, so a rep replying to a ticket already assigned and discussed internally can ask the Assistant to fold in that context automatically. Whether that produces a better reply than one written from scratch is not something the documentation measures; what it does establish is that the internal chat layer, previously visible only to teammates, now feeds directly into a drafted external message.

• Has an admin reviewed the connected AI provider's data-retention policy for the kind of customer information already flowing through shared threads?
• Would the 25 percent credit markup or a direct provider key work out cheaper at the team's actual monthly volume?
• What happens to in-progress drafts if the credit balance reaches zero mid-shift?

Missive's documentation treats internal thread context as a feature to be surfaced deliberately, not a default the Assistant reaches for without being asked, and treats the AI provider relationship as an organization-level decision rather than a personal setting. Both are worth confirming before a team assumes the Assistant already knows everything a thread contains.