Four channels,
one inbox
The chat on your website, WhatsApp Business, Instagram and Facebook Messenger all land on the same screen. Agents answer from one place regardless of the channel, with the customer record, past contacts and order details sitting next to the conversation. Outside working hours the conversation does not stop: CALLAII answers and hands over to the right agent when needed.
Connected channels
All four land in the same inbox. You do not need a separate team, a separate screen or a separate report per channel.
Website chat
A chat window added to your site with a single line of code. Which page the visitor wrote from, the campaign link and the session history are all in front of the agent.
WhatsApp Business
An official WhatsApp Business connection. Template messages, media and delivery status are supported; your business number stays the same.
Direct messages and post replies land in the same inbox. A sales question from social media does not get mixed up with support requests; it is tagged separately.
Facebook Messenger
Messages through your page arrive in one stream. When the same customer writes from a different channel, the history merges on one record.
Where the AI steps in
The rule is simple: a person speaks with the customer wherever possible, the AI fills the gap and keeps the door to a human open.
During working hours a person answers
While an agent is at the console the agent runs the conversation. CALLAII does not interrupt; it waits in the background and assists the agent when needed.
CALLAII answers when no agent is available
At night, on weekends or during a rush, a visitor is not left unanswered. CALLAII answers from your own knowledge base; another company's data never mixes in.
A handover is a real handover
When the customer asks for a person or the topic exceeds the limits, the conversation goes to an agent who is on shift and at the console right then. If nobody is available the task drops into the pool and the customer gets a promise of a reply, which stays tracked.
No bot on the phone
On a voice call a human agent always speaks with the customer. The hybrid model works in writing, not on the phone; that is a deliberate choice.
Gates that are not left to the model
Wherever AI speaks with a customer, the real risk is the model saying the wrong thing. That is why the critical decisions are made deterministically in code, without asking the model.
- Profanity, insults and threats: the response is applied before the model, escalating from a warning to closing the conversation.
- Out of scope topics: religion, politics, requests to write code and entertainment are politely refused without ever reaching the model.
- Company boundary: answers come only from that company's own knowledge base; tenant data never mixes.
- Closing discipline: a closing note cannot be skipped, gaps are found by query and the agent gets a warning.
Written conversations are measured too
Whatever exists on the call side exists in writing as well. The work does not end when the chat closes.
Chat analysis
A closed chat is evaluated with the same logic as a call: summary, sentiment reading, quality score and a risky conversation alert.
Live coaching
While the agent is in the conversation, a risky word warning appears instantly, produced without asking the model. The sentiment strip of customer messages updates while the conversation is still running.
The closing note flows into the CRM
The closing note is matched to the customer by phone or email and lands on the Customer 360 timeline. Where no match exists the note is not orphaned; it is kept on record.
Command center
Volume, sentiment distribution, satisfaction, agent status and channel breakdown on one screen. Which channel is backing up is visible without waiting for a report.
How this differs from classic live chat tools
Chat does not sit apart from calls
When the same customer writes first and calls later, both contacts merge on one record. In most tools chat stays in its own world with no link to the phone.
The bot is a switch you can turn off
AI replies are switched on or off per company. A company that does not want it uses the product as a classic live chat and loses nothing.
An engine trained for Turkish
The model that writes the replies was trained for Turkish. The tone and meaning errors that general purpose multilingual bots make in Turkish drop noticeably here.
Data in the EU region
Conversations are held in the EU region in a GDPR compliant way (and KVKK compliant for Turkey). Where nothing may leave the building, the Edge deployment is on the table.
Frequently asked
I do not want AI speaking with my customers. Can I turn it off?
You can. AI replies are a per company switch; when it is off the system behaves like a classic live chat and messages arriving out of hours wait on the agent screen. Even when it is on, it only steps in when no agent is available.
Does our WhatsApp number change?
It does not. An official WhatsApp Business connection is set up and your business number stays the same. After the connection, messages collect in the team's shared inbox in the panel rather than on a single phone.
We are worried the bot will give wrong information.
Answers are produced from the company's own knowledge base; it does not bring in general knowledge from outside. Out of scope topics are refused without reaching the model, and profanity or threats are handled deterministically. The moment it sees it has reached its limit the conversation goes to a person; handing over is preferred to making something up.
Can a handed over conversation get lost?
It cannot. A handover opens a task; if the assigned agent does not touch it within a short window the task drops into the pool and another agent picks it up. The promise made to the customer stays tracked, and an overdue promise lands on the manager's screen.
Can we get reports from the conversations?
You can. Every closed chat produces a summary, a sentiment reading and a quality score; the command center shows volume, satisfaction, agent status and channel breakdown in real time. The closing note is written onto the customer record in the CRM.
Let us see it on your own site
In the demo we show the chat window styled like your own site: what happens when a visitor writes, what happens when no agent is there, and how the handover works.