Home  ›  AI Agent
✦ Trainable AI Agent

An AI customer agent that speaks from your knowledge

Add your text, website, FAQs and files; the AI agent replies only from your knowledge (RAG). It runs in automatic, semi-automatic or suggestion mode — and hands off to a human when unsure.

AI Ajan · otomatik yanıt
FlexInbox AI
Bilgi tabanı + GPT-5
Kargo ne zaman gelir?
09:41
✦ AI Ajan
Siparişiniz 1-3 iş günü içinde teslim edilir. Kargo takip linki SMS ile gönderilir 📦
09:41 ✓✓

What is RAG, and why does it matter?

It's not a generic chatbot; it's an agent that answers from your data.

RAG (Retrieval-Augmented Generation) means the AI retrieves relevant pieces from your knowledge base before producing an answer. So the agent doesn't make things up; it conveys your products, prices, shipping and return rules from the source you wrote.

You feed the knowledge base by pasting text, crawling your website, adding FAQ question-answer pairs or uploading files (PDF, DOCX, CSV). When you update the content, the agent instantly replies with the current information.

  • Multiple sources: Text, website (a single page or the whole site), FAQs and file upload.
  • Personality & tone: You define the agent's personality, goal and rules.
  • Model selection: Choose from the GPT-5, GPT-5 mini/nano and GPT-4.1/4o families.
  • Playground: Try the agent in the panel before going live.

Three operating modes

Keep full control in your hands.

🟢

Automatic

The agent replies on its own; uninterrupted service even after hours.

🟡

Semi-automatic

Replies automatically above a confidence threshold; below it, offers a suggestion.

🔵

Suggestion

The agent drafts a reply, the operator approves and sends it — full control.

AI Ajan · otomatik yanıt
FlexInbox AI
Bilgi tabanı + GPT-5
Kargo ne zaman gelir?
09:41
✦ AI Ajan
Siparişiniz 1-3 iş günü içinde teslim edilir. Kargo takip linki SMS ile gönderilir 📦
09:41 ✓✓

In perfect sync with humans

The AI and your team work in the same conversation without clashing.

🤝

Automatic handover

The agent hands off to a human when unsure, or on keywords like "agent/complaint".

🤐

Operator priority

When the operator types from the panel or the phone, the AI automatically goes quiet in that chat.

📚

Learning from feedback

Give replies a 👍/👎 or write the correct answer; the correction is added to the knowledge base as an FAQ.

Credit-based, transparent usage

AI usage runs on a credit wallet. Each reply spends credits based on the model used and the token count — a pricier model spends more, a smaller one less. When credits run out, the agent stops and hands the conversation to a human; no surprise bills.

You can top up the wallet manually or set up auto top-up. You see exactly how many credits each message spent in the panel.

AI agent — FAQ

Will the AI make up answers?

The agent is fed only from your knowledge base via RAG and can be directed to hand off when no source exists. You can still review and correct every reply from the reports, making it more accurate over time.

Which channels does it work on?

Directly on WhatsApp and Telegram; through Bitrix24 Open Channels it works across all connected channels such as WhatsApp, Telegram, Instagram, Facebook and web chat.

Can I assign an agent to specific lines?

Yes. You can assign each line (number/Open Channel) to a separate AI agent, with a different personality and knowledge base for different brands/teams.

How many messages does 20,000 credits send?

It depends on the model you choose. A short reply (~700 tokens) costs about 14 credits with GPT-5, ~3 with GPT-5 mini and ~1 with nano — so 20,000 credits equals thousands of replies on the mini model.

Bring all your customer conversations into one panel

WhatsApp, Telegram, CRM and an AI agent — start free today, no credit card required.

Start Free →