The first channel

Deep on outbound voice — by design, not by limitation.

Real-time voice is the most demanding surface a communication engine can run. By mastering it end to end — pacing, concurrency, local-time windows, live transfers, and durable resume mid-run — we prove the whole loop. Outbound voice is the proof. The automation platform is the product.

Channel Outbound voice Telephony Bring your own Enforcement In the execution path

Why voice first

The hardest real-time channel proves the engine.

Asynchronous messages forgive a slow or flaky core. A live phone call does not. Voice forces every part of the platform to be correct at once — which is exactly why we built on it first.

01

Pacing under load

Calls are placed at a controlled rate with retry policies, not dumped onto the network. The engine paces itself against your limits rather than racing them.

02

A live concurrency ceiling

A tenant-wide ceiling caps how many calls run at once. The engine respects it continuously, in real time, while many automations compete for capacity.

03

Per-contact local time

Each contact carries its own clock. A run that spans time zones reaches each person inside their permitted local-time window — decided at the moment of dialing.

04

Transfers to people

A flow can hand a live caller to a real person mid-call. The handoff is part of the logic, not a dead end — and the result is recorded as an outcome.

05

Durable resume mid-run

Workflow execution is durable. If a process restarts, in-flight work resumes rather than restarting or vanishing — the run continues where it left off.

06

The full loop, proven

Build an audience, design the logic, launch a governed run, read what it achieved. Voice exercises every stage of that loop under real-time pressure.

Treat outbound voice as the proof that the engine works. The control plane from audience to insights is what the platform is.

What voice does today

Spoken flows, IVR, capture, and transfers — all shipped.

Every capability below runs over the customer’s own telephony today, designed as steps in the visual flow builder.

Available today

Shipped voice steps

Multi-voice, multi-language text-to-speech

Speak generated audio in a range of voices and languages, composed with templated {{ expressions }} from per-contact data.

IVR menus

Present a menu, route on the caller’s choice, and drive the rest of the flow from what they selected.

Keypad / DTMF capture

Collect digits the caller enters — confirmations, account numbers, menu choices — and store them against the contact.

Surveys

Ask a sequence of questions, capture responses, and turn structured answers into recorded results.

Conditional branching and delays

Branch on captured data or expression logic, and insert governed delays between steps to shape timing.

Transfers to live people

Hand a call to a real person as a defined step in the flow, then record how the conversation resolved.

Audio playback from a library

Play pre-recorded audio from your in-app library alongside generated speech.

Record results

Map what happened on each call to declarative business outcomes that feed analytics — not just call logs.

AI in voice — what’s real, what’s deferred

We are precise about AI, because the category is trust.

There is real AI in voice today, and there is AI we have deliberately not shipped. We will not blur the line between them. The product does not include a real-time conversational AI voice agent today — and that is a deliberate choice, not an omission we are hiding.

A

Today

Available today

  • Natural, multi-language AI-generated speech for what the flow says.
  • A live expression engine with templated {{ expressions }} that personalize speech and drive branching from per-contact data.

This is deterministic, inspectable behavior — the flow does exactly what its steps and expressions define.

B

On the path

On the path

  • An AI node category — AI speech, sentiment and intent analysis — is reserved in the product surface but not yet available.
  • A real-time conversational AI voice loop, where the system understands a caller and responds live, is deliberately deferred. We are not faking it.
  • AI agents and AI-driven workflows that branch on understanding sit ahead of us on the roadmap, not in today’s build.

Reserving a node in the surface is a commitment to direction, not a claim of capability. When the conversational loop ships, it will ship as a labelled capability — not as something quietly switched on.

Governed by default

Every call clears the gate before it dials.

In voice, governance is not a report you read afterward. A pre-dial compliance gate runs in the execution path, before any number is called, and the engine honors live pause, resume, and stop while a run is in flight.

Suppression / DNC
Numbers on your suppression and do-not-contact lists are blocked before dialing.
Local-time window
Each contact is checked against its own permitted local-time window at the moment of the call.
Per-contact frequency caps
Contact attempts are held within the caps you set per contact.
Concurrency ceiling
A tenant-wide live ceiling caps simultaneous calls across all running automations.
Live pause / resume / stop
Operator controls are honored by the engine mid-run, not queued for later.
Durable resume
If the system restarts, in-flight runs resume rather than restart or drop.

See how enforcement and isolation work in Trust

Pre-dial gate — illustrative Run #4417
+1 415 ··· 0192Cleared
+1 312 ··· 7741Blocked · DNC
+1 206 ··· 3380Held · outside local window
+1 617 ··· 5026Cleared
+1 720 ··· 1184Held · frequency cap
12 in flight ceiling 20

These are tools to operate within your own rules. The platform enforces the controls you configure — it does not make you compliant, and it claims no certification on your behalf.

From voice to many channels

One model, built to span more than voice.

The same flow logic, the same declarative outcomes, and the same in-path governance are designed to carry across channels — not rebuilt per channel. Voice proves the model; the model is meant to generalize.

On the path

Channel surfaces

Voice Available today

Outbound voice over your own telephony, fully governed and outcome-mapped — the channel we master first.

SMS On the path

SMS is reserved in the product surface and not available today. It is direction, not a current capability.

Email is not a channel surface today. It exists only as a step you can place inside a flow — already part of the same logic model, but not a standalone channel peer to voice. A second channel surface is on the path, not shipped.

Read where the multi-channel platform is headed

  Start with the channel that proves the rest

See outbound voice run end to end.

Bring your own telephony, design a flow, and watch a governed run reach the right contacts at the right time — with every call clearing the gate first.

Watch a demo Explore the platform