Bland shipped a change on August 3, 2026 that removes a setting most AI calling teams have been tuning by hand. The resumption_speed field is gone. Agents now measure each caller's pace and adjust their own response timing automatically.
If you run outbound call campaigns on Bland, this changes what you tune in a pathway, and what you should stop tuning entirely. Here is what shipped and how we are handling it on client builds.
What Bland shipped
Two changes in the August 3, 2026 changelog. Bland notes the release is rolling out now and may not be active in every environment yet.
Adaptive resumption
Resumption is what happens after the caller stops talking: how quickly the agent takes its turn. Bland now measures two things per caller, how long their speech runs and how long their mid-thought pauses last, and waits accordingly.
- Fast callers get near-instant responses.
- Deliberate callers get an agent that waits out a thinking pause instead of stepping on it.
- A caller who changes pace mid-call is matched within about fifteen seconds.
- It runs in both directions: slowing down earns patience, speeding back up restores snappy responses.
The resumption_speed field is removed, and response timing no longer has a manual setting. Bland's own line on configuration: "Nothing to configure."
Node-scoped interruptibility
On pathways, a per-node interruptibility setting now applies only while that node is active. Leaving a node that declares a setting reverts to the agent's own setting instead of carrying the node's value for the rest of the call. The same now applies to per-node interruption thresholds.
Two controls, not one
Worth keeping straight, because the update touches both. Per Bland's agent speech docs, interruptibility governs what the agent does while it is speaking and the caller talks over it. That is still yours to set: an integer from 0 to 4, default 2.
- 4 Always interrupt
- 3 Easy
- 2 Balanced (default)
- 1 Difficult
- 0 Block interruptions
You set it on the /v1/calls request body for a whole call, on an inbound number configuration, or per node in the pathway node inspector under Advanced options. Resumption, what happens after the caller finishes, is the part that is now automatic.
Why this matters for outbound teams
Three reasons.
It was the wrong knob to expose. Response timing was a per-agent global you had to guess at. Guess fast and the agent talks over a prospect who was mid-thought, which is the exact moment a cold call dies. Guess slow and it feels laggy to the decisive buyer who was about to say yes. There is no single correct value because there is no single caller.
Call cohorts vary more than anyone accounts for. We run outbound calls across real estate, healthcare, and professional services. A busy broker at 8am talks in fast clipped bursts. A practice manager reading off a screen pauses mid-sentence. Same script, same agent, two completely different pacing requirements. A static setting served one of them and annoyed the other.
Node-scoped interruptibility makes long pitches safe. The old behavior punished you for using the setting at all. Turn interruptibility down for one node, say a pricing explanation or a compliance disclosure, and the agent stayed stubborn for the rest of the call, including the parts where you want it to yield instantly. That made a useful setting a liability. Now it is scoped, so you can turn it down exactly where it earns its keep and get default behavior back the moment you leave that node.
Our read: this is a maintenance win more than a conversion win. It deletes a class of tuning that never really converged, and it removes a footgun that punished correct use of a per-node setting.
How we'd use it
1. Strip resumption_speed out of everything that still sends it. The field is removed. Check your call payloads, your n8n workflows, and any Clay-to-Bland handoff. Do this before you debug timing behavior, or you will lose an afternoon chasing a setting that no longer exists.
2. Re-audit per-node interruptibility now that it is scoped. Walk every pathway and look at each node with a non-default value. Under the old behavior, some of those were set to compensate for a value bleeding forward from an earlier node. That compensation is now wrong. Our standard on pathway builds: balanced (2) everywhere by default, drop to 1 or 0 only on nodes where the agent must finish (a phone number readback, a price, a required disclosure), and go to 3 on nodes where you expect the prospect to push back.
3. Stop tuning timing, start tuning the script. When an agent feels off on a call, the cause is rarely the gap between turns. It is an opener that sounds like a script, a question that traps a busy person into a long answer, or an objection branch that does not exist. That is where your recorded-call review time should go now.
4. Re-baseline before you judge it. Bland says this is rolling out and may not be live for you yet. Note the date it lands on your account, then compare connect-to-conversation rate and booked meetings against the two weeks prior. Hold the list and the offer constant while you measure, otherwise you will credit this update for something your targeting did.
FAQ
Do I need to change my Bland agent settings after this update?
Two things. Remove resumption_speed from anything that still sends it, since the field is gone. Then review per-node interruptibility on your pathways, because a node's setting no longer carries past that node.
What is the difference between interruptibility and resumption in Bland?
Interruptibility is how readily the agent yields while it is speaking and the caller talks over it. You set that, 0 to 4, default 2. Resumption is how quickly the agent takes its turn after the caller stops. That is now automatic and measured per caller.
Why does my Bland agent still interrupt callers?
Interruptibility is a separate control and adaptive resumption does not touch it. If the agent cuts people off while it is talking, lower interruptibility on that node. If it jumps in too quickly after they finish, that is resumption, and Bland's measurement catches up to a caller's pace within roughly fifteen seconds.
Get your calling system built right
AnaqVisual builds and runs outbound systems end to end: cold email infrastructure, AI calling agents on Bland and Vapi, and Clay-built lists feeding both. We tune pathways like these for clients every week, which is why an update this small still gets a same-day audit here. Choosing a platform? Start with our Bland AI vs Vapi breakdown.