Workflow 01 of 05 · Front desk

Your front desk, answered.
After hours first.

Calls, texts, WhatsApp and web chat answered from rules you can read, so every customer gets the same answer and none of it goes to a model. Start it on the calls you miss today: after hours, overflow, the lunch rush. Widen it when you trust it.

Hear it answer, live: +1 248-509-0888

A real Feerasta receptionist line. Call it, interrupt it, ask for a human. It runs on the same pipeline you would get.

Watch it answer

One thread, three rules.

What a customer sees at 8:42pm, and what each reply was made from. The thread is an illustration; the rules are the ones in the dental template you can read on day one.

  1. 1
    It says what it is

    On the first turn of a call or text it says it is AI and names your business. You can turn that line off; the liability for doing so is yours.

  2. 2
    Hours and new patients come from a rule you can read

    Hours, prices, whether you are taking new patients, the cost of a check-up: plain conditions and plain replies, edited by you, with a trace that shows which rule matched.

  3. 3
    Pain goes to a person, whatever you configured

    Pain, swelling, anything clinical, anything about a bill, a complaint, or a request for a human is handed off on every channel: transferred to the number you set, or a message taken. It does not guess and it does not hang up.

What it replaces

The line item it takes off your desk.

Each card is the thing you pay for or do by hand today, and what this workflow does instead. The price is set against that line item rather than against software seats.

Replaces: Scripted humans in a call centre

The answering service

A front desk that knows your hours, your prices and your booking rules, and says the same thing at 2am as at 2pm. Anything it was not taught, it takes a message for instead of guessing.

Replaces: Voicemail and phone tag

The unanswered phone

Calls, SMS, WhatsApp and web chat land in one thread per customer. A contact found mid-sentence becomes a lead you can see, with the whole exchange attached.

Replaces: A cloud model improvising on your behalf

The risky chatbot

No model call in the customer's path. Replies come from your approved workflow and FAQ through one resolver, which is why it is free per conversation and never invents a policy.

How it proves itself

It has to earn the switch.

The same rule on every workflow: it reproduces your own past before it runs, a person approves what it drafts, and the log keeps every decision. Every channel starts switched off, and a channel with a failing backtest of your own history cannot go live. The six controls, and what we have not earned yet →

1

Rules you can read

Start from one of eight industry templates or write your own. Every rule is plain language: when a customer says this, do that, and every run produces a trace including the rules that did not match.

2

A safety battery before go-live

Seven standards, four of them critical: regulated advice, complaints, a request for a human, and refusing instructions hidden in a message. Sixteen fixed hostile probes plus up to twelve generated from your own FAQ. A critical failure blocks activation.

3

Every conversation scored

In production, 100% of eligible exchanges are audited, not a sample. Findings go to a person; a confirmed failure becomes a permanent probe for every future activation.

Day one, draft-first

What you actually see.

You start with drafts and traces rather than an empty screen, no card is taken, and every decision stays yours.

Minute 5
Pick a template for your industry and read the rules it comes with: hours, prices, booking, and what always goes to a person.
Minute 15
Test it in the console. Type what your customers actually ask and watch the trace. Edit a rule, run it again.
Hour 1
Run the safety battery. Nothing goes in front of a customer while a critical standard fails.
Day 1
Draft-first on the web widget. Replies queue for your approval until you decide the channel is ready to answer on its own.
When you are ready
Point after-hours and overflow calls at it. Staff-only first, then public. Your hours run on your timezone, and a caller who asks for a person is transferred to the number you set, or leaves a message if you have not set one.
Draft first, a person approves The path an outbound action takes: a trigger fires, the agent writes a draft that never sends, and an approver approves, edits or rejects it. An approved draft then has to clear four gates in order, a consent record on that channel, an armed channel, the calling window in your own timezone, and the AI disclosure line. All four passing sends the message; any one failing writes the refusal and its reason to the audit log. FOUR GATES, BEFORE ANYTHING LEAVES EVENT DRAFT APPROVED ANY GATE FAILS ALL FOUR PASS A trigger fires missed call · invoice due The agent drafts it never sends Your approver decides approve · edit · reject GATE 1 Consent on record per person, per channel STOP revokes every one GATE 2 The channel is armed every outbound channel starts switched off GATE 3 Inside your hours read in your timezone, not ours GATE 4 Disclosure attached says it is AI and names your business Audit log the refusal and the reason Sent to the customer with the disclosure line LEGEND the one place a person decides refused, and written down the log you can export

The same path for every outbound action, whatever the workflow. A draft is written, a person on your team decides, and four gates stand between that decision and anything leaving.

Scroll the diagram sideways to see the whole path.

Honest limits

What it does not do yet.

Written down so you are not surprised. The roadmap is public in how we prove it.

Pricing

Flat. Unlimited seats. The price is the price.

No credits, no per-message meters, no success fees. Natural-voice minutes are billed at cost. Every tier, in your currency →

Free

$0
USD

One workflow, draft-only, no outbound sends. Enough to build your front desk and test it against real questions.

Team

from $149
USD / month

Live channels, owner and approver roles with email invites, consent ledger, data export, monthly verification report. Flat by workflow count, unlimited seats.

Front desk

Try it on your own questions.

Draft-first from the first minute. Nothing sends, pays or files until your team says so, and your data is never used to train a shared model.

Call the demoBook a call