How Rely works / AI voice agents

The voice agent performs only approved actions.

The agent performs tested, approved actions, discloses that it is AI, and stops at configured boundaries.

Map a transition Safety and governance

Define every agent stop and accepting owner before launch.

The agent moves through tested actions until a configured signal changes the owner.

Scope

Approved for this workflow, not in general

Every action is approved for one workflow and switched off everywhere else.

Place and take calls
Answer bounded questions from the approved set
Collect approved information
Schedule or coordinate where permitted
Document the disposition

A stop, drawn

A configured stop routes the case to a named owner.

The edge of scope

**A stop is designed behaviour, not a failure.** The agent works up to the line drawn before launch, stops there, and routes the thread over it to a named owner. Clinical decisions sit past that line and stay with your licensed team.

The whole record

**Every attempt is on the case, not only the one that connected.** Outreach runs the cadence agreed for the program, and each attempt carries its channel, time and outcome into the record.

A stop becomes an owned transfer

The agent stops for clinical questions, urgent or safety-sensitive language, prohibited actions, untested languages, identity uncertainty, or a failed dependency.

Context collected before the stop
A named destination
Explicit acceptance by the next owner
A fallback when that owner is unavailable

Inspect what happens after the agent stops.

A useful demo shows the owner, context, acceptance, and fallback after the agent stops.

Map a transition See a workflow

The voice agent performs only approved actions.

Define every agent stop and accepting owner before launch.

Inspect what happens after the agent stops.