Taking over and resuming: the handoff workflow
When your agent flags a conversation, you can take over under your own name or resume the agent with a context note it reads on its next reply.
Updated June 26, 2026
When the Needs Attention queue shows a
paused conversation, you have three choices: take it over, resume the
agent with a note, or ignore it. This article covers when to use each
and how the agent actually receives your guidance.
Where handoffs happen
Any row on the Needs Attention page that represents a paused
agent (sentiment stop, keyword stop, transfer_to_human call,
manual pause, etc.) shows two buttons:
- Resume agent — green. Hand control back to the agent. The
- Take over — outline. Pause the agent under your name and
Both open the same modal. The modal also surfaces the humanised
pause reason at the top — "Hostile sentiment," "Keyword matched
(stop, unsubscribe)," "Agent asked for help," etc. — so you know
what you're deciding on before you click.
Resume agent — the common path
Most "flagged" conversations don't actually need a human to step
in. They need a human to look at the flag, decide it's fine, and
hand it back — sometimes with context the agent should carry
forward.
Example — hostile sentiment false positive:
Contact texts: "That's the worst thing I've ever heard. I love it."
The SENTIMENT stop condition fires on
the word "worst." The agent pauses. You look at it, realise they
were being sarcastic, and click Resume agent. You type in the
modal:
> "They're being sarcastic — they love the product. Keep going.
> No handholding needed."
Click Resume. The agent unpauses and sees your note on its very
next reply. It continues the conversation with full context.
Two ways to resume — wait vs send now
The Resume modal has a "Send a follow-up message now" checkbox.
This is the key control that decides what happens immediately
after Resume:
- Unchecked (default) — wait for the next inbound. The agent
- Checked — agent sends a follow-up now. The agent composes
If working hours are enabled on the
agent and the current time is outside the window, the follow-up
is skipped (the agent is still unpaused; the note is still
saved). The modal tells you this so you can choose whether to
wait for the next window or disable working hours.
When to leave "Send now" off
- The contact was mid-sentence when the pause fired — they'll
- You just took over for 5 minutes, you're handing back before
- You want the agent ready-but-quiet
When to tick "Send now"
- The conversation has been idle for hours and you want momentum
- You finished a phone call with the contact and the agent
- The contact never messaged back after the stop-condition hit
Take over — when the agent shouldn't come back alone
Use Take over when you want to handle this conversation
manually and not leave it up to the agent to pick up again
automatically.
Example — angry contact threatening a lawyer:
Contact says "I'm going to call my attorney." SENTIMENT fires.
You click Take over with the reason:
> "Handling directly, will update when resolved."
The conversation stays paused. You reply to the contact yourself
from the Inbox. When you're done, return to Needs Attention
(or the contact page) and click Resume agent — this time with a
note telling the agent what happened:
> "Contact and I reached agreement on the refund. They'll wait
> 7 days for the credit to clear. Confirm receipt once they
> respond. Do not bring up the complaint again."
The agent resumes, reads the handoff note, and picks up from the
new baseline.
How the agent sees your note
When you resume with a note, it's stored in the contact's memory
under a category called handoff_context. That entry gets
injected into the agent's system prompt on every subsequent turn
under the heading "What You Already Know About This Contact":
## What You Already Know About This Contact
- handoff_context: A human just handed this conversation back to
you. Their note: "Contact and I reached agreement on the refund.
They'll wait 7 days for the credit to clear. Confirm receipt
once they respond. Do not bring up the complaint again."
Treat this as essential context for your next reply; do not
re-open topics the human has already addressed.
The agent reads this alongside your system prompt, persona, and
knowledge base on every turn. Instructions like "don't re-open the
complaint" reliably stick because the agent sees them fresh each
time.
What makes a good handoff note
- Summarise what the human leg of the conversation covered.
- Tell the agent what NOT to do. "Don't re-ask about their
- Note any time-sensitive commitments. "They'll follow up in
- Don't dump customer PII you didn't get permission for. The
Takeover without a follow-up note
If you take over and then decide you don't want to resume the agent
at all (e.g. you converted the lead, you closed the complaint,
they're going to deal with it on their end), leave the conversation
paused. It drops off the Needs Attention queue once the
conversation stalls or the contact marks it done — nothing else
happens. The agent won't spontaneously reply.
You can also apply an opt-out tag via rules or
manually in GHL to permanently stop outbound to that contact.
Finding the handoff history
Every takeover and resume writes an audit trail entry with the
operator's identity, the note they left, and the timestamp. Visible
in two places:
- Audit log (sidebar → Audit Log) — workspace-wide timeline
- Contact detail page — per-contact history with the same notes
Useful for post-mortems ("why did we drop this lead?") and for
onboarding new operators ("this is how the team handles hostile
contacts").
Keyboard-friendly workflow
- Click Resume agent on the row
- Type the note (autofocus is on the textarea)
- Enter submits (via the form's default-submit behaviour)
- Row disappears from the queue
- Sidebar badge decrements by 1
Most handoffs should take under 15 seconds end to end.
The badge on the sidebar
The red count next to Needs Attention in the left nav is the
number of rows currently on the queue. It polls every 30 seconds
and decrements when you resolve a row. If the number keeps
climbing, that's usually a sign of:
- A stop condition firing too
- An agent too quick to call
transfer_to_human(tighten its - A real external change — e.g. a bad batch of leads that are all
Treat a rising badge as a signal to investigate, not just to
clear.