New Flow Builder and Knowledge Base are live. What changed

Product

A message comes in. Six stops later it reaches an outcome.

Describing Chatinbox with a feature list would be misleading; the product is not a list, it is a flow. Below we follow a single customer message from start to finish — with the panel as it looks at each stop.

Short answer

This page describes Chatinbox not as a feature list but as the six stops a single customer message passes through: arriving from a channel, landing in front of the team, an AI suggestion, a person sending it, repetition turned into a rule, and a record you can find later. The example is from a clinic; the same stops work in e-commerce. At each stop you see the panel as it looks at that moment.

The conversation in the example is from a clinic; the same six stops work the same way for e-commerce sellers too — only the content of the message changes.

The most-asked question first

The AI prepares the suggestion; your team sends it. In the automated flows you build, the bot sends the messages you defined without waiting for approval. These are two separate layers and we write down both — the full list of where we stand.

The 15-minute intro call is free and carries no commitment. We handle the setup; the setup fee depends on the scope of your business and appears as a separate line in the written quote.

How to read this page

Every capability on the site is sorted into one of five states. The flow below describes only the ones that are available today; the rest sit in a separate list with their own label.

  • Available today — in the product, customers use it.
  • Coming soon — written, but not open in the panel.
  • Awaiting confirmation — we could not find the evidence, so we do not write it.
  • Roadmap — planned, does not exist today.
  • Not available — we do not do it, and we say so.

The screens below are illustrative drawings of the Chatinbox interface. The example conversation is illustrative too; it contains no real patient data.

A message arrives

Three channels, one queue

Zeynep · WhatsApp · 09:10 Hello, how much is a hair transplant? My hairline is thinning.

WhatsApp Business API, Instagram DM and cloud PBX (call) conversations collect in a single panel; website live chat (WebChat) is coming soon. Telegram and Facebook Messenger can also be connected — we set them up for any customer who asks. Whichever channel the customer writes from, your team looks in one place; the question "did they write on WhatsApp or Instagram?" disappears.

Channels and integrations →

Clinic inbox: seven prospective-patient conversations in one list, with waiting times and reminder tags visible
Clinic inbox

It gets read

A messy message lands in front of the team in order

AI reading: intent "general enquiry", urgency low, purchase intent present.

When a long or tangled conversation arrives, the AI summarises it, classifies what is being asked and leaves a reading of how serious the request looks. None of this goes to the customer — all of it stays on the team's screen.

  • A summary of a long conversation
  • What the customer is asking for
  • How serious each conversation looks
AI reading card: sentiment, urgency, customer requests, suggested action and conversation summary
The reading card the agent sees

The source is checked

The suggestion is fed by your own knowledge

The clinic's price document feeds the suggestion; the agent gets three tones at once.

The content you upload to the knowledge base is given to the AI as the source for its suggested reply. Alongside it, the saved replies your team approved earlier come into play. The source feeds the suggestion; it does not limit it — a suggestion is produced even for a question the knowledge base has no answer for. What catches a weak suggestion is the next step: the agent.

  • The suggestion is fed by your own documents
  • Repeated answers live in one place
Reply suggestion card: a suggestion fed from the knowledge base, three alternative tones and the sources behind it
A suggestion fed from the knowledge base

A person decides

The person is the one who sends it

The draft lands in the agent's compose area. It is a person who presses "Reply".

The AI suggests a reply to the agent; the decision to send it to the customer belongs to the agent. This is the most-asked part of the product and the one with the clearest answer — but only for the suggestion layer: In the flows you build, the bot sends the messages you defined to the customer without waiting for your approval. These are two separate layers and we write down both.

The full list of where we stand →

The agent's compose area and Reply button — a person starts the send
The send decision sits with the agent

Repetition becomes a rule

Do not do the same job by hand twice

If the photo never arrives, the conversation should not be forgotten: the reminder you set lands in front of the agent.

Repeating steps such as out-of-hours greetings, tagging and routing can be set up as rules. Setting a rule does not mean the AI decides on its own: you write the rule, the system applies it.

  • Repeated steps become rules
  • Repeated answers live in one place
Reminder card inside the conversation: if the photo has not arrived, come back for consultation eligibility
The trace a set reminder leaves in the conversation

A trace remains

When someone looks back, there is an answer

In the record: new prospect, note added — "Asking about Sapphire FUE price and graft count."

You can see which agent wrote what to which customer, which requests are still open and which conversations are waiting for follow-up. The conversation is linked to the customer record, and someone without permission cannot see it.

  • Who wrote what to whom, and what is still open
  • The conversation and the customer record sit together
  • Each person sees only what they are authorised to see
Prospective-patient record: prospect tag, conversation note, channel and access information with the conversation trace on one screen
The conversation turned into a record

The same path, uninterrupted

This is how the six stops above look one after another in the panel. 43 seconds, no voice-over.

Recorded in a demo environment; the example conversation is illustrative and contains no real patient data.

Let's review your message flow together in 15 minutes

Let's talk about what's scattered across which channel, and see together how your team would use Chatinbox.

We handle the setupWe train your teamYour WhatsApp number stays the same

ChatinboxChatinbox

A WhatsApp, Instagram DM and call workspace for clinics and e-commerce sellers. AI suggests; your team sends the draft.

Ask an AI about Chatinbox

G C P

© Chatinbox 2026 · Chatinbox Yazılım Arge ve Teknoloji Limited Şirketi · chatinbox.net

Chatinbox chatinbox

Looking into Chatinbox? I answer setup, pricing and channel questions right here.