Skip to content
ATLASBIP Business Intelligence Partners
ATLASBIP Business Intelligence Partners

CRM · AI SaaS Live product

Cavably AI CRM for service businesses with a shared WhatsApp and Instagram inbox

Five channels. One inbox. Answers at any hour.

Cavably is our AI-first CRM for salons, clinics, courses and other service businesses. We designed, built and run it. Messages from every channel land in one inbox, an AI assistant answers from the business’s own knowledge, and a conversation turns into a booking, a payment or a deal without leaving the screen.

Project facts

Project type
Our platform
Platforms
Web app · Web chat · WhatsApp · Instagram · Messenger · Telegram
Languages
AZ · EN · RU

23:40 The challenge

Customers write at midnight. The front desk opens at ten.

A service business lives in its messengers. Questions, bookings and promises to pay arrive on WhatsApp, Instagram, Telegram and the website, at any hour and on several phones. Whether anyone answers depends on who is holding which phone.

One day of messages at a sample studio Opening hours Answered while open Waits until morning
Opening hours 00:00 06:00 12:00 18:00 24:00 WhatsApp Instagram Messenger Telegram Web chat Nərmin · 23:40 00:00 06:00 12:00 18:00 24:00
  1. The message nobody saw

    A booking request arrives on Instagram at 23:40. By the time the studio opens, the customer has booked with someone who answered.

  2. One slot, two promises

    An admin confirms 15:00 on WhatsApp while a colleague gives the same slot to a caller. Both customers arrive.

  3. Debts in a notebook

    Who paid a deposit and who still owes for the last visit lives in someone’s memory, a notebook or a spreadsheet nobody updates.

  4. The fiftieth price question

    Prices, the address, parking and opening hours are typed by hand all day, by the people who should be serving customers.

  5. No idea what works

    Which post, ad or link brought a customer is gone the moment the chat scrolls away.

23:40 Shared inbox

Every channel lands in one inbox

WhatsApp, Instagram, Messenger, Telegram and the website chat arrive in one shared inbox. Every conversation has an owner, a full history and a customer profile, and the whole team sees in real time who is answering what.

  • Routing by workload

    Rules hand each new conversation to the right operator. Anyone already at capacity is skipped, so nobody drowns while a colleague waits.

  • No double replies

    Typing and “already replying” signals appear the moment a colleague opens the same chat.

  • Quick replies with “/”

    Saved answers drop into the chat with a slash, and their variables fill in with the customer’s own details.

  • Follow-ups that go out on their own

    An unanswered chat gets a nudge: free text inside WhatsApp’s 24-hour window, an approved template after it.

  • Comments that turn into chats

    A keyword under an Instagram post triggers a public reply and a private message with the details.

  • Every lead with its source

    Tracked links, QR codes and click-to-WhatsApp ads tag each new conversation with where it came from.

23:41 AI assistant

An assistant that knows the price list and knows when to stop

The AI assistant replies at once in Azerbaijani, Russian or English. It answers only from what the business has taught it, understands voice notes and photos, and passes the conversation to a person the moment it should.

  1. 1 Knowledge base

    Answers from the business’s own knowledge

    The team adds FAQs, free text and pages of its website. Cavably splits and indexes them automatically, and the assistant grounds every answer in those sources.

    • FAQ · 24 answers
    • Price list
    • Website · Services page
  2. 2 Voice notes

    Hears voice notes

    Customers often record instead of typing. The assistant transcribes the voice note, shows the transcript in the conversation and answers. It can also reply with a voice note of its own.

  3. 3 Photos

    Understands photos

    A customer sends a screenshot or replies to a Story. The assistant recognises what is in the picture and finds the closest match in the business’s catalogue.

  4. 4 Handoff

    Hands over at the right moment

    When the knowledge base does not confidently cover a question, such as a complaint or a medical concern, the conversation goes to a person with the whole history attached. The business decides how readily the assistant hands off.

23:41 No-code flow builder

Automations drawn, not coded

Owners and managers draw their own conversation logic on a canvas: triggers, questions, conditions, waits, AI steps and handoffs. A built-in simulator runs a sample message through the flow before it is published.

  • Every building block

    Text and buttons, media, lists, WhatsApp templates, questions saved to variables, if/else, A/B splits, waits and business-hours branches.

  • AI inside the flow

    An AI step classifies what the customer wants and picks the branch, answers from the knowledge base, or keeps the chat with the assistant.

  • Connected to the rest

    CRM actions, handoff to an operator, HTTP requests, Google Sheets rows, email and other flows are nodes on the same canvas.

23:42 Bookings and payments

The chat becomes a booking. The booking gets paid.

A time agreed in a conversation goes straight into the right specialist’s calendar. Clashes are caught before they happen, confirmations and reminders go out on their own, and money owed is tracked next to the visit.

  • No double bookings

    Each specialist has a calendar. A time that is already taken is refused before anyone confirms it.

  • Confirmations and reminders

    The customer gets an automatic confirmation when the visit is booked and a reminder before it.

  • Money next to the visit

    Payments and balances are recorded against the customer, so the admin sees who owes what before the visit starts.

  • Debt reminders on the due date

    Add a payment with a due date, and a polite reminder goes out on that day without anyone remembering to send it.

Day 60 Pipeline, campaigns and loyalty

Conversations become deals. Customers come back.

Larger sales like a course, a treatment package or a corporate order move through a pipeline that starts from the conversation itself. Regular customers collect points, and a customer who has gone quiet gets a well-timed message instead of being forgotten.

  • A pipeline born in the chat

    A deal is created from the conversation and stays linked to it. Stages, won and lost reasons, pipeline value and forecast are one click away.

  • Campaigns that respect consent

    Broadcasts go only to customers who opted in, use approved WhatsApp templates, and drop anyone who writes STOP.

  • Loyalty points and reactivation

    Points for every visit, and automatic win-back messages when a regular customer goes quiet.

Mon 09:00 Analytics

Monday morning, the whole week on one screen

The owner sees how conversations turn into bookings and money: the lead funnel, the revenue trend, which channel brings which customers, how fast the team replies and how satisfied customers are after each chat.

  • From message to money

    The funnel follows each lead from the first message to a booking and a payment.

  • Satisfaction after every chat

    A short rating request goes out when a conversation closes. A low score alerts the manager and reopens the chat.

  • Channels and sources compared

    See which channel, link, QR code or ad brings conversations, and how quickly each one is answered.

Engineering and security

Enterprise controls inside a small-business CRM

Cavably is a multi-tenant SaaS built to satisfy a company’s IT team as well as a salon owner. Single sign-on, an append-only audit log, SLA policies and custom roles sit next to the inbox.

How it is built

Multi-tenant
One platform, many businesses: each company works in its own tenant, and its data is isolated from every other.
Real time
New messages, typing, presence and “already replying” signals reach every operator’s screen the moment they happen.
Grounded AI
Retrieval over each business’s own sources, with a handoff threshold and a monthly AI usage limit that the business sets itself.
Security
Role-based permissions, Google sign-in, SAML 2.0 and OIDC single sign-on, an append-only audit log, and channel keys stored encrypted.
Open
Webhooks, an open API with scoped keys, and ready connectors such as Google Sheets.
Three languages
Azerbaijani, Russian and English across the interface and the AI; each business sets its language, time zone and working hours.

Connected channels and systems

Messaging channels

  • WhatsApp
  • Instagram
  • Messenger
  • Telegram
  • Web chat
  • Click-to-WhatsApp ads
  • Google Sheets
  • Webhooks
  • Open API
  • Email
  • Google sign-in
  • SAML 2.0 / OIDC

Systems and sign-in

Our role

What Atlas BIP did

  1. Product and UX

    We shaped Cavably around the working day of a salon, clinic or course: the inbox first, and bookings, payments and deals one click from the conversation.

  2. AI engineering

    Retrieval over each business’s knowledge, intent routing, voice transcription and voice replies, photo understanding and confidence-based handoff.

  3. Channel integrations

    WhatsApp, Instagram, Messenger, Telegram and an embeddable web chat, with templates, broadcasts, opt-outs and comment-to-message rules.

  4. Platform engineering

    A multi-tenant, real-time SaaS where the inbox, bookings, payments, pipeline, campaigns, flow builder and analytics share one data model.

  5. Enterprise security

    Single sign-on with SAML 2.0 and OIDC, an append-only audit log, SLA policies and custom roles with fine-grained permissions.

  6. Launch and operation

    We localised the product and its AI in three languages, launched it at cavably.com and run it as our own platform.

Stack

What Cavably runs on

Frontend
  • TypeScript
  • React
  • Tailwind CSS
Backend
  • Node.js
  • NestJS
  • WebSockets
Data
  • PostgreSQL
  • Redis
AI
  • LLM APIs
  • RAG
  • Speech-to-text

Questions and answers

What buyers ask about Cavably

Is Cavably a real product I can open today?

Yes. Cavably is our own platform, live at cavably.com. We designed it, built it and run it, from the AI assistant and channel integrations to the enterprise controls.

How does the AI avoid making things up?

It answers only from the sources the business adds: FAQs, text and website pages. When an answer is not confidently covered, or the customer asks for a person, the chat goes to an operator with the full history. The business sets how readily that happens.

Which channels and systems does it connect?

WhatsApp, Instagram, Messenger, Telegram and a web chat widget on the business’s site, plus click-to-WhatsApp ads, Google Sheets, webhooks, an open API, Google sign-in and SAML 2.0 or OIDC single sign-on.

Does it work in Azerbaijani?

Yes. Azerbaijani is the default language, and the full interface and the AI assistant also work in Russian and English.

Can you build an AI assistant or messaging CRM like this for our company?

Yes. We build the same pieces for other businesses: knowledge-base AI with handoff, voice and photo understanding, WhatsApp and Instagram integrations, bookings, pipelines and no-code automations, either inside your existing system or as a new product.

Your project

Your system is next

Tell us how orders, money and documents move through your company today. We reply within 24 hours with questions and a concrete plan.

Names and figures on screens are sample data.

Back to the start