Pattern 3
Let callers opt into the agent from the queue they're already in
A press-2 option inside the wait message you already play. Callers who choose the agent arrive more willing to work with it, and anything it can't finish sends them back to a person at the front of the queue.
The shape of it
What changes and what stays
The option
Press 2
Added to the wait message you already play
Everyone else
Same queue
Unchanged for a caller who does nothing
If it can't finish
Front of queue
Not the back. Opting in costs nothing
Then
Overflow
The same option covers the queue at peak
Self-selection
Why an opt-in caller is a better first tester
A caller who reaches the agent because a rule sent them there never chose it. A caller who presses 2 did, and that changes how the same rough edge reads: a routed caller reads it as a broken line, an opt-in caller reads it as a new product they tried.
Businesses that outsource overflow to a second contact centre buy patience they can't measure or improve. An opt-in agent buys the same patience from your own callers, and every call it takes lands in your own transcript store, scored the same way as any other call.
Explainer
The pattern, explained
Where this breaks
The mistake is turning press 2 into a default
The easy fix on a bad day is to make the agent the default during a spike, so every caller in the queue gets it whether they pressed 2 or not. That is a different pattern with a different risk profile, and doing it without deciding to is how a trusted opt-in option turns into a forced one overnight.
Keep the two separate. An opt-in extension proves the agent works before you ask it to cover the queue outright. Collapsing the two steps skips the proof.
The pattern
Five steps
01
Add the option to the wait message you already play
One new line: the current wait time, then the choice. Press 2 to talk to the AI agent now instead of holding. Nothing else about the message changes.
02
Route press 2 through the extension already in your dialer
No new number, no new channel. The DTMF digit maps to the agent the same way any other extension on the PBX does.
03
Send anything the agent can't finish to the front of the queue
Not the back. A caller who tried the agent and still needs a person does not pay a second wait for having tried.
04
Score every opt-in call against scenarios before trusting the volume
The caller chose this call, so the pass rate on it is the cleanest signal that the agent is ready for calls nobody opted into (docs.meddle.sh/docs/operate/evals).
05
Widen from there
The same option that took the first opt-in call takes the queue at peak once the scenarios hold. That's the overflow pattern, reached without a forced rollout.
The payoff
The same option on the queue at peak
Once the scenarios hold, the opt-in extension is the smaller of the two benefits. The larger one is the same as the out-of-hours pattern: an agent proven on the calls that chose it can be forwarded the queue at peak, covering the calls that would otherwise wait past your target. The maths behind that peak is on the queue maths page, and the forwarding options are on inbound calls.
On Meddle
Building it
Wiring
One extension, no new channel
The press-2 digit maps to the agent through the dialer extension already in your phone system. The build is a few minutes of configuration once the agent itself is ready.
Return path
Front of queue, not the back
A call the agent can't finish transfers back to a human at the front of the queue. The caller who tried the agent never pays a second wait for it.
Proof
Scenarios and a judge
Write the opt-in query as a scenario, run it, and read the grade. The scenario runs again on every change to the agent (docs.meddle.sh/docs/operate/evals).
After
Every opt-in call in the transcript
Each call is stored turn by turn with timings and scores, and you choose the retention window. Opt-in volume is the cleanest read you get on whether the agent is ready for more.
Last updated 14 September 2026
Questions
Before you add the option
Does the press-2 option replace the existing IVR?
What happens if the agent can't finish the call?
Why do opt-in callers make a better first test group than a routed caller would?
Does this need a new phone number or channel?
Can the opt-in run during business hours, not only after hours?
Add the option to your wait message.
Build the agent, wire the extension, and script the first opt-in query as a scenario. Turn it on once the graded calls hold.