
Alder 36" oak vanity
$1,180
The problem

People ask in the evening and on weekends, when nobody is at the desk.

The phone rings during a job or a meeting, and the caller gets voicemail.

By morning, a lead who waited all night has often booked someone else.
44%
of calls to businesses never reach a person.
Invoca, Lead Conversion Benchmarks Report 2026. More than 70 million calls.
77%
of customers expect to reach someone right away when they contact a company.
Salesforce, State of the Connected Customer, 2023. 14,300 people in 25 countries.
The solution
Website chat
Every visitor gets a real answer at a person's pace, from your pages, files and price list. Step into any chat, then hand it back.

Website chat
11:38 PMPhone line
It picks up every call with the same knowledge as the chat and passes the call to a person when asked.

Incoming call
(512) 555-0178Leads and follow-up
Each chat and call becomes a lead with a score, a short summary and its source, down to the ad click. The follow-up runs on its own.

Follow-up - running
After a missed callBuilt for
Pick the job. The Operator sets up the rep, the channels and the tools for it, and you change anything you like.

SaaS companies
It answers from your docs, looks up each user's own account through your API, and opens a ticket when your team is needed.

Businesses that live on leads
The chat on your site and a phone number answer at any hour, ask your questions, and book the visit before the lead goes elsewhere.
Coming soonOnline stores
Connect your Shopify store. It answers about products, sizes and stock, and tells customers where their order is, without a ticket.
The Operator

Operator
Tell the Operator what you sell and what a good lead looks like. It reads your site and files, and writes the rep, its greeting and its questions. No prompts, no code.

Tests - running
Before go-liveThe Operator runs tests on the questions your customers ask, and grades every answer. Each change waits for your Apply, and Undo is one click.

Overview - last 30 days
412
Chats
96
Leads
31
Booked
Chats a day, by week
Chats, leads and bookings by day, page and ad. Ask the Operator why a number moved, and it reads the conversations and shows you.

Calendly, Slack, Twilio and Linear today, plus webhooks, a REST API, and your own API or MCP server as a tool.
For technical teams
Everything the Operator sets is yours to read and change by hand. Nothing is locked in a black box.

01
Instructions
Plain Markdown with fields like {{company}}. Every save is a version you can restore.
02
Models
11 models from OpenAI, Anthropic and Google, a backup on another provider, and the reasoning effort.
03
Custom tools
Your own API or MCP server as a tool, signed for each user, so it can act on their account.
04
Webhooks and API
lead.created, call.booked, ticket.created and more, plus a REST API for leads and chats.
05
Tests and replay
Must and must-not rules that run on every save. Replay any real reply against a new version.
06
Auto rollback
If a new version brings in fewer leads, the version before comes back on its own.
Capabilities

It learns your site, price lists and FAQs, and answers only from them. Change a price once, and every answer follows.
It checks your Calendly and books the slot inside the chat, in the visitor's time zone. You get a booking, not a to-do.
Before a reply goes out, each price and date in it is checked against your sources. A wrong one is caught and rewritten.
Connect your own API or MCP server. It looks up the customer's account, opens the ticket in Linear and tells your team in Slack.
Choose from 11 models by OpenAI, Anthropic and Google. If one is slow or down, a backup on another provider answers.
Your rules
A front desk may want a rep with a name and a person's pace. A product or a store may want a clear AI agent that shows its steps. You choose for each rep, and Operator suggests one.

Nora
Active now
Nora joined the conversation
NoraJust now
A name, a face and a quick rep's pace, in your tone. For a front desk: your homepage, a local business.

Prices and dates are checked before they are sent. When it does not know, it says so and takes the visitor's details.

Take over any chat, quietly or by name, then hand it back. On a call it transfers to your number.

Ask it what your customers ask on a private page, and go live only when you like the answers.

Each save is a version. Flag a reply you would say differently, and the fix is tested before it ships.


On the offer page, day and night

“It answers hard questions with real client results, day or night, and sends ready buyers to checkout. It only says what we taught it.”
Sarah Evans
CEO, Zen Media
Pricing
No plans, no seats and no monthly fee. Your rep uses credit as it answers.
Per chat reply, about
A chat is two or three replies.
Per phone minute, about
With the transcript and a summary.
Per member, agent or subaccount
Add as many as you need.
Work out your month
Estimates. Receptionist: U.S. Bureau of Labor Statistics median pay, May 2024. Fin prices from fin.ai/pricing, October 2026, with Fin on your current helpdesk. Fin bills only chats that end in an outcome; this example assumes 60% do, and Fin lists no phone price. Our price depends on the model and the length of each chat. Fin and Intercom are trademarks of Intercom, Inc.
Security and reliability
Your rep talks to your customers, so we treat their data as yours. Here is how it is kept safe and running.
Every model runs through the business APIs of OpenAI, Anthropic and Google, which do not train on what your customers say.
Each draft is checked before it is sent. An amount or a date that is not in your sources is caught and rewritten.
Each answer keeps a step-by-step record: the sources it searched, the model that wrote it, the checks it passed and what it cost.
If an AI provider slows down or goes offline, a backup from another provider answers. Your customers never notice.
Before any automated call, text or email: opt-outs, consent, contact limits and the person's local hours are checked.
TLS in transit and AES-256 at rest. Each account is sealed off at the database level, not only in the app.