An AI customer agent is software that answers customer questions like a human, but instantly and 24/7. In FlexInbox, this agent speaks only from your knowledge (RAG).
An AI customer agent (AI chatbot) is software that understands incoming customer messages and produces answers in natural language. Unlike classic "button bots", it also answers freely typed questions. The agent in FlexInbox runs on a RAG (Retrieval-Augmented Generation) model: before producing an answer, it retrieves relevant information from your knowledge base, so it doesn't make things up; it conveys your product, price, shipping and return rules from the source you wrote.
This agent works directly on WhatsApp and Telegram; and through Bitrix24 Open Channels it works on all connected channels including web chat, Instagram and Facebook.
Including after hours, the customer gets an answer without waiting; your first-response time approaches zero.
Thanks to RAG it speaks only from your knowledge base — the risk of generic/made-up answers drops.
The AI takes on routine questions; your team focuses on value-adding work.
When unsure, or on words like "agent/complaint", it hands off to a real person automatically.
You give replies a 👍/👎 or write the correct answer; the correction is added to the knowledge base.
Credit-based usage; when credits run out, the AI stops — no surprise bills.
Add text, your website, FAQs and files; set the agent's personality and rules.
Choose which WhatsApp/Telegram number or Open Channel is answered by which agent.
Choose automatic, semi-automatic or suggestion (the operator approves) mode; load credits into the wallet.
With RAG it is fed only from your knowledge base; when no source exists it can be directed to hand off to a human. You can review and correct every reply from the reports.
The model is multilingual; it can reply in whatever language you write the knowledge base in and in the language the customer writes.
No. When the operator types from the panel or the phone, the AI automatically goes quiet in that chat; control is always yours.
It depends on the model: a short reply (~700 tokens) costs ~14 credits with GPT-5, ~3 with GPT-5 mini and ~1 with nano — so thousands of replies on the mini model.
WhatsApp, Telegram, CRM and an AI agent — start free today, no credit card required.
Start Free →