Does a suggestion send by itself?

No. It lands in the reply box for a person to edit. Automatic sending is a separate feature, controlled by the workspace AI switch.

Where do suggestions get their facts?

From the same connected sources the automatic agent uses: knowledge base passages, the customer's orders, your catalog, your hours and the thread so far.

Can new team members use it on day one?

Yes. A suggestion encodes what is in your knowledge base, so a new agent starts from your answer.

What is the difference between a canned reply and a macro?

A canned reply is message text you send. A macro changes the state of the conversation, such as tagging, noting, assigning or closing.

Can I edit canned replies in the app?

They are picked while you write. The library itself is not edited in the interface.

Do ratings change future suggestions?

Ratings are stored for review. What actually changes the next suggestion is editing the knowledge base the suggestion reads from.

Does a scheduled reply skip the usual checks?

No. It is rebuilt at send time and passes the same consent, conversation window and duplicate checks a live reply does.

Can two agents draft on the same thread?

A draft lock warns when someone else is already composing, so two people do not answer the same customer twice.

Customer conversations

Start from a draft, not a blank box.

The words are already there. You decide which ones survive.

Agent assist is the set of tools that speed up a human agent without taking the reply out of their hands. A suggestion is written on request and lands in the reply box for editing, common replies are saved for reuse, and repetitive housekeeping is bundled into one click. The person still decides what is sent.

Ask for a suggestion, then edit it before it goes
Rate a suggestion so the good and bad ones are recorded
Canned replies for the answers you send constantly
Macros that tag, note, assign and set status together

How it works.

01
Open the thread
Read the customer's message with the contact profile and past conversations beside it, the same context the AI agent would get.
02
Ask for a suggestion
Ask for a draft. It is written from your knowledge base and the customer's own history, and it lands where you type instead of going out.
03
Edit and send
Change the wording, cut a line, or replace it entirely. Nothing leaves until you press send, and the send follows the same consent and window rules as any reply.
04
Rate it
Mark the suggestion useful or not. Ratings are stored against the conversation so the quality of assistance can be reviewed.

Suggestions on demand

A suggestion is generated when an agent asks for one. It is grounded in the same sources the automatic agent uses, which keeps a human reply and an AI reply consistent with each other.

It lands where you type, and never sends on its own
Built from your own answers, your order data, and your hours
Rated up or down and recorded for review

Saved replies, a keystroke away

The answers you send twenty times a day live as reusable replies, scoped to the channel they suit, and carry a risk marker for the ones that need care.

Channel-scoped so a template does not appear where it cannot send
Risk and approval markers on the sensitive ones
Picked while you type, without leaving the thread

Macros for the housekeeping

The repetitive part of closing a conversation is rarely the reply. A macro does the tagging, noting and assigning in one click.

Add a tag, add a note, set status, assign
Built-in follow-up and escalation macros to start from
Your own macros for how your team actually works

Summaries for handovers

AI written recaps are stored as internal messages in the thread, so an agent picking up a long conversation can read the shape of it before the detail.

Recaps stored in the thread as team-only messages
Visible to the workspace, never sent to the customer
Sits alongside the notes your team writes by hand

Send-later and reminders

Assistance includes deciding when, not only what. A reply can be scheduled and a thread can be brought back to your attention at a set time.

Scheduled replies rebuilt and re-checked at send time
Personal reminders that reopen the thread and post a note
Consent, the reply window, and duplicates all re-checked at the moment it sends

What it does not do.

Saved replies are picked while you write. The library itself is not edited in the interface.
Suggestions are written on request. They do not appear on every incoming message.
Ratings are recorded for review. They do not retrain a model on their own.

Common questions

Does a suggestion send by itself?: No. It lands in the reply box for a person to edit. Automatic sending is a separate feature, controlled by the workspace AI switch.
Where do suggestions get their facts?: From the same connected sources the automatic agent uses: knowledge base passages, the customer's orders, your catalog, your hours and the thread so far.
Can new team members use it on day one?: Yes. A suggestion encodes what is in your knowledge base, so a new agent starts from your answer.
What is the difference between a canned reply and a macro?: A canned reply is message text you send. A macro changes the state of the conversation, such as tagging, noting, assigning or closing.
Can I edit canned replies in the app?: They are picked while you write. The library itself is not edited in the interface.
Do ratings change future suggestions?: Ratings are stored for review. What actually changes the next suggestion is editing the knowledge base the suggestion reads from.
Does a scheduled reply skip the usual checks?: No. It is rebuilt at send time and passes the same consent, conversation window and duplicate checks a live reply does.
Can two agents draft on the same thread?: A draft lock warns when someone else is already composing, so two people do not answer the same customer twice.

Every question in one place, answered once.