Relay runs your outbound phone floor. You train an AI agent in plain sentences, load your list, and set the hours it is allowed to dial — and every call comes back with its transcript, its recording and an outcome your team can work from.
EU voice stackTenant isolation in the databaseEnglish · German · Albanian
For teams that sell by phone — across every industry that does
Three decisions shape everything else in the product.
Behaviour is written as rules a manager would recognise: always do this, never say that, take this step, answer this when they ask, handle this objection. Relay compiles them into one instruction for every call, and every call records the agent that ran it.
When the agent finds somebody worth calling back, it hands the lead back with the transcript, the outcome and the reason attached. It waits in a single follow-up queue until someone claims it.
Calling windows are a floor the dialler cannot cross — enforced in the API, in the dial path, and in the query that reserves the next call. Outbound stays off on a new account until an operator turns it on.
What you sell, who you call, how the agent behaves, who owns the lead — all in one place, per account.
The price, the monthly payment, who it is for, what you pitch and the objections you keep hearing — written once on the product and compiled into every call.
Personas with a voice per language, guardrails on what they may never do, and rehearsals you can run before a single stranger is dialled.
Import your customers with the fields you actually use, assign them to supervisors, and control exactly which of those fields the agent is allowed to say out loud.
Split shifts and overnight windows, a per-campaign outcome vocabulary, and retry rules — with a compliance floor that outranks all of it.
Every call with its transcript, its outcome and its recording — and every hot lead the agent handed back, sitting in a queue waiting to be claimed.
Volume, connection rate and conversion by day and by supervisor — next to exactly what the AI consumed, so the cost of a booked appointment is never a guess.
Add the product, the price, who it is for and what you pitch. Add the objections you already know are coming.
Brief it the way you would brief a new hire, or have it drafted for you and edit what comes back. Publish — and if the script needs a detail your list does not carry, Relay refuses and tells you which field, and what to do.
Import your customers, choose the windows the agent may dial in, and pick the outcomes that count as a win for this campaign.
Follow live calls as they happen, claim the hot leads it hands back, and read the transcript and outcome of everything else the next morning.
Four parts of Relay, close up — each shaped by a real day on a phone floor.
The persona carries a distinct voice for English, German and Albanian — chosen per language, not run through a translation layer. The greeting, the outcome vocabulary and the small talk are native to each.
You write the call the way you would brief a new hire, in your own words, on the product. Every call compiles it into one instruction — and if it uses a field your list does not carry, the save refuses and names the field.
Operators run accounts, admins run campaigns, supervisors watch the calls. Every account is isolated at the database level, and the calling window is enforced wherever a dial could start.
One call, stage by stage, exactly as the board shows it — ringing, live transcript, outcome, follow-up card.
Supervision in Relay is read-only. You follow live calls and their transcripts as they happen, and you work what the agent hands back afterwards — the console has no way to speak into a call in flight, and the mic below never turns on.
Not demo tricks — the everyday machinery of a phone floor, built in.
Every hot lead the agent hands back waits in a single queue until someone claims it.
The audio, the transcript and the outcome, kept together on the call.
The next call is reserved atomically — two agents never dial the same person.
Split shifts, overnight windows, per campaign — enforced below the UI, in the very query that reserves the call.
“Take me off your list” flags the customer on the spot — and the flag outranks every campaign.
Each campaign defines what counts as a win; every call ends tagged with one.
No-answers come back on the schedule you set — and stop when the rules say stop.
From the first dial to the win — every stage counted from calls the system actually made, never from a spreadsheet someone maintains by hand.
Illustrative funnel — example figures, not customer data.
Outbound calling is regulated, and a demo that ignores that is not a product. These are properties of the system, not promises in a policy document.
Let the agent work the top of the list — the dials that never get made — and keep your closers on the conversations that are actually worth their hour.
Every client is a separate account with its own products, agents, hours and numbers, isolated from the others, all under one operator console you control.
Stand a client up in an afternoon: their offer, their voice, their language — without writing a prompt or shipping a line of code.
The voice, the greeting and the outcome vocabulary are chosen per language — not a translation layer bolted on top of an English agent. Your supervisors read the same screens in the language they think in.
Pick one to read this page in it.
No price list to decode — we scope the plan to your call volume and seats on one short call. These are the shapes it comes in.
For a first campaign.
For a whole phone floor.
Everything in Pilot, plus
For call centres and resellers.
Everything in Floor, plus
Already running campaigns? Sign in to your console.
Tell us what you sell and who you call. Leave your details and we will call you back — you'll hear the agent first and a human second, which is rather the point.