What is Response Eagerness?
When a user finishes speaking, the agent needs to decide when to respond. This requires determining whether the user has actually finished their turn or is simply pausing mid-thought. Getting this timing wrong has consequences:- Too early — The agent interrupts the user, cutting off their thought and creating a frustrating experience
- Too late — The conversation feels sluggish and unnatural, with awkward silences after the user finishes speaking
Eagerness Levels
Start every step on Keen and extend to Normal or Patient only where callers are slower to respond. New steps are created on Keen. A step with no eagerness saved is treated as Normal.
Set it from the step’s menu: Response Eagerness → Keen, Normal or Patient.
How It Works
When to Use Each Level
Keen (Default)
Start every step on Keen. It suits nearly every exchange: questions, confirmations, acknowledgements and ordinary conversation. The assistant still waits longer when the caller’s sentence sounds unfinished (see Turn Completion Detection below).Normal
Move a step to Normal only when the caller is slower to respond because the line asks them to find or check something before answering.Patient
Move a step to Patient when the answer is read out slowly in pieces — a name, a phone or account number, a reference code, an email address or anything spelled out — so a pause between chunks isn’t taken as the end of the answer.Turn Completion Detection
Voxworks doesn’t just use timers. It also analyzes whether the user has completed their thought:
This prevents interrupting users mid-thought, even with faster eagerness settings.
Per-Step Configuration
Eagerness is applied step by step. Every time the assistant speaks, the listening for that step starts fresh with that step’s own eagerness setting — so a keen step really is keen, even if the step before it was patient. Set eagerness per step to match the expected interaction:Balancing Speed and Accuracy
New steps start on Keen. Adjust based on testing.
Interaction with Other Settings
Eagerness works alongside other dynamics:
To lengthen or shorten the wait after the caller stops speaking for the rest of a call, a Code Step can call
endpoint_offset(seconds). The wait never drops below the built-in minimum.
Best Practices
- Start on Keen — New steps start on Keen, and that is the right setting for almost every step. Move a step to Normal only where the caller has to look something up first, or to Patient where the answer is a name, number or code read out in pieces
- Configure per step — Set eagerness on each step based on expected response complexity
- Extend only where callers are slower — Keep Keen unless testing shows the assistant jumping in while the caller is still finding or reading out their answer
- Test with real conversations — Timing feels different on a live call
- Consider your users — Some audiences prefer more deliberate pacing
Next Steps
- Thinking Effort — Control LLM reasoning depth
- Silence Tolerance — Handle idle users
- Fast Response — Decide whether a step waits for a reply
- Overview — See all conversation dynamics