Skip to main content

​
Why this is a common failure point

Apple and Google both ship call screening that answers first, asks who is calling and why, and then decides whether to put the call through, decline it, or take a message. An outbound agent that doesn’t recognise a screener delivers its opening pitch to a robot. The call may even connect afterwards, so the outcome looks fine while the person has already missed the introduction.

​
How Roark tests it

Call screening dials your agent into real screener behaviour, one call per screener outcome, and grades whether the agent noticed it was talking to a screener, answered it directly with who is calling and why, and held its opening for the person.

​
Setting it up

Pick the screeners and outcomes to call into. Each screener offers only the outcomes that product actually has: One call runs per outcome, per agent. For a walkthrough, see Testing call screeners.

​
What it measures

The template seeds these, and you can add or remove metrics in Advanced before running.

​
What to look for

  • The first agent turn. A pitch delivered to the screener is the most common failure, and it happens in the first few seconds.
  • Declines and messages. An agent that argues with a decline, or leaves no message when offered, fails differently from one that misses the screener.

​
Over the API

Run it with POST /v1/simulation/run and "template": "call-screening". GET /v1/simulation/template lists what each template accepts.