Tasks and approvals
Your Tasks page is a queue to empty, not a dashboard to read. Your AI Agents draft outreach overnight and stop there; nothing reaches anyone until you send it.
The page has two tabs.
Messages to send
One draft at a time, with everything you need to judge it beside it. No opening the contact, no opening the company, no hunting for the research.
For each draft you can:
- I've sent this manually — you copied the message and sent it yourself. This logs it so your AI Agents can follow the thread and never draft it again.
- Send to LinkedIn — appears once your LinkedIn is connected. Doers sends it from your account, inside your working hours.
- AI Refine — sitting under the message itself. Tell your AI Agent what to change and it rewrites the draft; you review the new version before anything goes out. The sparkle button beside it hands the message over with no instructions at all, and your AI Agent reviews it on its own.
- Skip — not now. Pick how long to wait (a week, two weeks, a month, or a date of your own) and this person leaves your queue until then. Your AI Agents won't draft for them in the meantime, and they come back on their own when the time is up. Nothing about them is lost. Press Ctrl + Enter (⌘ + Enter on a Mac) to confirm without reaching for the mouse.
Two less common controls:
- Start over — the refresh icon above the message. After you confirm, it throws this draft away and has your agent write a fresh one. If the research behind it is out of date (older than 14 days, unless your team changed it in Automatic AI work), that is refreshed first — unless your team turned refreshing off there. Your spacing between messages still applies; if it's too soon you'll see when the next message is allowed, with a Write anyway option.
- Do not contact — in the ⋯ menu next to Skip. After you confirm, it stops all outreach to this person, for good. Reach for Skip when it is only a matter of timing.
A draft leaves this queue on its own once a newer message reaches that person — for example when you log a message you sent yourself, or send one from LinkedIn directly. There's nothing stale to clean up.
You can also select any part of the message to ask your agent about just that sentence.
What the chips tell you
Along the top of each draft:
- Who it's for — their name opens the contact, the company chip opens the company, and the LinkedIn icon opens their profile.
- Channel — for email, the address is shown in full with a copy button, so you can paste it straight into your inbox.
- Angle — the customer problem this message leads with.
- First touch (green) or Follow-up (amber) — an opener and a follow-up read completely differently, so this is colour-coded.
Why this problem? and Why this message? open your agent's reasoning if you want to check its thinking. Why this message? names the move it made — for example "New way into the same problem" or "Proof, inline" — and why it chose that move for this person right now.
Two ways to work
- List — the whole queue down the left, so you can jump to a particular person first. This is how the page opens.
- Focus — one draft at a time, with nothing else on screen. Best for clearing a queue quickly.
The keyboard is faster
Clearing a full queue is much quicker from the keyboard. Press ? on the page for the full list:
| Key | Does |
|---|---|
| Shift + S | Mark the current draft as sent |
| → | Next draft, without resolving this one |
| ← | Previous draft |
| E | Edit the message |
| I | AI Refine — tell your AI Agent what to change |
| K | Skip this person for a while |
| C | Copy the message |
Shortcuts pause while you're typing, so you can't fire one by accident. Marking a message sent needs Shift held down as well — it's the one key here you can't take back, so it takes a deliberate press.
Your edits are saved, and your AI Agents learn from them
Press E (or the pencil) and the message becomes editable. Change whatever you like, then press Save changes — your version becomes the message, and it stays yours if you come back to it later.
This is worth doing even when you are about to send anyway. Your AI Agents read the wording you settled on and draft more like you next time, so the drafts you get tomorrow need less fixing than the ones you get today. Cancel throws your changes away and brings back the version you started from.
If you would rather send straight away, you can: the send buttons stay available while you are editing, and sending saves your wording at the same time.
Research, without leaving the draft
Beside every message: the AI Agent chat (ask for changes and watch the draft update), Contact Research, Company Research, and every message already exchanged with this person.
Decisions
Your AI Agent keeps learning about you, your ICPs and your company. When it wants to change something it knows, it asks here first, so nothing about your targeting or your writing style changes behind your back. Each person sees the changes that come from their own work: the drafts they reviewed, the contacts they own and the notes they wrote. A teammate's learnings stay in their queue, so you only ever act on what you know about. Admins can switch to a teammate's queue, the same way as for messages; see Whose queue is this?.
Read the list at a glance
Changes are grouped by where they land: Your profile, each of your ICPs by name, Problems, Companies & contacts, Your company, and Other actions. Each row says what changes (for example "Update Personal tone") and, underneath, exactly where:
- The coloured dot is the persona. Two ICPs can both have a "CTO"; the dot and the ICP card tell them apart.
- in … is the group the information belongs to, such as "Writing style".
- Affects … appears on problems: a problem can be shared by several personas, so one change reaches all of them.
Use the check to approve and the cross to dismiss straight from the list, or open the row to look closer.
Review a change
Opening a row shows the change itself: new lines with a green bar, removed lines struck through with a red bar, and small wording edits with the changed words highlighted. Parts that stay the same are folded away. Show final text shows how it will read once saved. The small i next to the title explains what that piece of information is used for.
Then choose:
- Approve saves it right away.
- Edit before approving (the pencil) lets you fix the wording in place, then Approve my version. Your AI Agent is told what you changed, so its next suggestion is closer to how you'd say it.
- Dismiss leaves everything as it is. You can pick a reason or write your own before Dismiss change. It's optional, but it's how your AI Agent learns not to suggest the same thing again.
Want the evidence first? When a change comes from what your AI Agent noticed in your work, a line under the change says what it is based on, for example Based on 3 meetings across 3 contacts, and each message, note or meeting behind it is one click away. Your AI Agent only suggests these once something repeats, so a one-off comment of yours doesn't turn into a rule. Why this change? gives its full reasoning; the small numbered badges in it, and in the change itself, open the source each fact came from. After you decide, Next takes you straight to the next one. For a step-by-step walkthrough with screenshots, see Review what your AI Agent wants to change.
If you see This changed since it was proposed, someone updated that information after your AI Agent made its suggestion. For a change to existing information you can still edit it, starting from today's text; new information someone already filled in can only be dismissed.
Questions from your AI Agent and other steps waiting for your go-ahead appear here too. Open them to answer, approve or dismiss.
Your own direct edits never wait for approval. Approvals only apply to changes your AI Agents propose.
Whose queue is this?
Each person sees the drafts written for them. See Who can see and do what for how that works across a team.