Every screen, and what it actually does.
Nine parts of the product, in the order you meet them. The panels on this page are drawn in the product’s own design language to show the shape of each screen. They are illustrations, not screenshots, and the data in them is made up.
It learns your business from the site you already wrote.
You paste your web address and it reads the pages the way a new hire would on their first morning: services, hours, locations, the FAQ you wrote three years ago and forgot about.
What it finds becomes a list of entries you can read, edit and delete. If your site is thin or out of date, this is where you find that out, and you can fix an entry in place rather than rewriting your website first.
- Nothing to type in to get started, and no scripts to draw
- Every entry is editable, so a wrong answer is one line to fix
- It answers from these entries, so it is not guessing at your policy
The one screen where a template becomes your agent.
Identity, greeting and behaviour in plain sentences, not prompt syntax. You describe how the agent should talk to callers the way you would brief a temp, and you get the same editor on every plan with nothing gated behind a tier.
Behaviour splits into two short lists on purpose. Hard rules are the things it must always do. Guardrails are the things it must never say. Short separate lines mean you can remove one rule without breaking the rest.
- Greeting you write, or have it stay silent until the caller speaks
- Hard rules like confirming a callback number before hanging up
- Guardrails like never quoting an exact price, on by default
- The same editor on Basic, Growth and white label
Eighty-plus voices, and you can hear every one.
Filter by gender, accent and language, play a sample, and pick the one that sounds like your front desk. Samples are pre-recorded, so listening to twenty of them costs you nothing.
The pick is not a commitment. Change the voice later in the agent editor and the next call answers in the new one, with nothing else to redo. Language is handled separately: whichever voice you choose, the agent follows a caller who switches language mid-call.
- 80+ built-in voices with samples you can play before choosing
- 25+ languages, detected automatically and switched mid-call
- Filters for gender, accent and language to narrow the list fast
- Change the voice later without touching anything else
It knows when to stop talking and get a person.
Add transfer contacts in priority order with a ring timeout on each. If the first person does not pick up in time it tries the next, and if nobody does it takes a message rather than dropping the caller.
You define the trigger conditions in plain language: the caller asks for a human, the topic is sensitive, the caller sounds frustrated. Whoever picks up hears a short summary first, so they are not starting the conversation from zero.
- A fallback chain, not a single number that might ring out
- Trigger conditions written the way you would say them out loud
- A spoken summary handed to whoever answers
- Falls back to a message with the summary if nobody is free
The paperwork happens without you.
Every action that runs after a call is a toggle with a small amount of configuration. Summary email to you, booking written into your calendar, contact logged into a sheet or your CRM, Slack message to the channel you name.
You choose which of these fire and on which outcomes. A caller who only wanted your hours does not need to generate four notifications.
- A summary email to you after every call, or only some
- Calendar booking created or updated in Google or Outlook
- Contact logged into Google Sheets or your CRM
- Slack notification scoped to the outcomes you care about
The caller hears back before they have put the phone down.
Separate from the summary that goes to you, the agent can send the caller an email of their own. A booking confirmation with a reschedule link, directions, or the information they asked about on the call.
Triggers are conditional rather than blanket. You pick which outcomes get a follow-up, and the wording is yours to edit in the same plain language as the rest of the agent.
- A confirmation on a booking, with your address and a way to reschedule
- A longer email when the caller asked for detail
- Fires only on the outcomes you choose, not on every call
- Templated by agent type, then edited by you
Every call, written down and kept.
A searchable list of calls with the number, the length and what the call ended in. Open one and you get the full transcript alongside the recording, filterable by agent, date and outcome.
Recordings and transcripts are served through authenticated endpoints tied to your workspace, never as public links that anyone with the address could open.
- Full transcript and audio for every call, exportable
- Filter by agent, date and outcome
- Served behind authentication, not from open URLs
- The record of what was actually said, when a caller disputes it
The numbers that tell you whether it is working.
Calls answered, bookings made, answer rate and average call length, per agent. Plus calls by hour, which is usually the first chart that changes an owner's mind, because it shows the spike that used to go unanswered.
Resellers get the same report per client, brandable, ready to forward at the end of the month.
- Calls answered, booked and transferred, per agent
- Average call length and answer rate
- Monthly report generated for you, brandable for partners
- Minutes used against your pool, so a bill is never a surprise
Ask about your calls. Change your agent by saying so.
An assistant inside the dashboard that reads your own call logs and agent configuration. Ask how the week went, or why a particular call did not end in a booking, and get an answer grounded in your data rather than a generic chart.
It also edits the agent. Describe the change and it proposes the exact rule or setting it would update, then waits for you to confirm. It does not quietly rewrite your configuration.
- Plain questions about your own calls, answered from your own logs
- Changes described in words, applied to the right field
- Every change shown and confirmed before it saves
- Scoped to your workspace only
Why it does not have the phone-bot pause.
You never see this layer and you never configure it. It is worth one section because it is the reason the calls sound the way they do.
One model, not three systems
Most phone bots convert speech to text, send the text to a language model, then read the reply back with a separate voice engine. Suvy Voice runs on a single speech to speech model, so the handoffs that create the pause are not there.
Failover is a configuration change
The primary voice engine can fail over to a compatible backup engine, so switching providers changes a configuration value rather than rewriting how your agent behaves. Your rules, voice and connectors are untouched by it.
Recordings and traces come from the engine
Call recordings, transcripts and traces are part of the voice engine rather than a feature bolted on afterwards, which is why they exist for every call on every plan instead of on a higher tier.
It writes into the tools you already open.
The five standard connections are included on every plan, at no extra cost.
Google Calendar
Reads your free slots on the call and writes the booking in.
Outlook Calendar
Same booking behaviour if your business runs on Microsoft.
Slack
Pings the channel you choose for the calls you want flagged.
Google Sheets
A row per caller, for businesses that never wanted a CRM.
CRM
Drops the contact and the call outcome into the CRM you already run.
Your own system
A custom connection into your booking software or point of sale.
Connecting your own booking software or point of sale costs extra on the direct plans, and is included on white label. See what that costs.
Reading about it is slower than using it.
Signup is self-serve, there is no setup call, and you get three test calls in the browser before a customer reaches it. From $49 a month with no setup fee.